Update architect priorities after A2A handoffs

.github/workflows/harness.yml

@@ -75,20 +75,6 @@ jobs:
             -summary-json provider-conformance-summary.json \
             -summary-markdown provider-conformance-summary.md \
             -capabilities-markdown provider-capabilities.md
-      - name: Publish provider conformance summary
-        if: always()
-        run: |
-          if [ -f provider-conformance-summary.md ]; then
-            cat provider-conformance-summary.md >> "$GITHUB_STEP_SUMMARY"
-          fi
-          if [ -f provider-capabilities.md ]; then
-            {
-              echo
-              echo "## Registered provider capabilities"
-              echo
-              cat provider-capabilities.md
-            } >> "$GITHUB_STEP_SUMMARY"
-          fi
       - name: Upload provider conformance summary
         if: always()
         uses: actions/upload-artifact@v4

agent/agent.go

@@ -207,6 +207,17 @@ func (a *agentImpl) Stream(ctx context.Context, message string) (ai.Stream, erro
 	})
 }
 
+// Resume returns the response for a checkpointed agent run. Completed runs are
+// returned from the checkpoint without calling the model or replaying tool
+// calls; failed or in-progress runs continue from the saved input message.
+func Resume(ctx context.Context, ag Agent, runID string) (*Response, error) {
+	a, ok := ag.(*agentImpl)
+	if !ok {
+		return nil, fmt.Errorf("agent resume: unsupported agent implementation %T", ag)
+	}
+	return a.resume(ctx, runID)
+}
+
 // Pending returns checkpointed agent runs that have not completed. It mirrors
 // flow.Pending for startup recovery loops that drain durable agent work.
 func Pending(ctx context.Context, ag Agent) ([]flow.Run, error) {
@@ -289,10 +300,6 @@ func (a *agentImpl) askLocked(ctx context.Context, runID, message, parentRunID s
 		run.Status = "paused"
 		run.State.Stage = agentApprovalStep
 		run.State.Data = []byte(message)
-		if a.pause.Tool == toolHumanInput {
-			run.State.Stage = agentInputStep
-			_ = run.State.Set(inputPause{OriginalMessage: message, Prompt: a.pause.Message})
-		}
 		run.Steps[0].Status = "paused"
 		run.Steps[0].Error = a.pause.Message
 		run.Steps[0].Result = a.pause.Tool

agent/builtin.go

@@ -20,9 +20,8 @@ import (
 // the discovered service tools. There is no separate harness or graph:
 // the LLM calls them like any other tool.
 const (
-	toolPlan       = "plan"
-	toolDelegate   = "delegate"
-	toolHumanInput = "request_input"
+	toolPlan     = "plan"
+	toolDelegate = "delegate"
 )
 
 // builtinTools returns the tool definitions exposed to the model in
@@ -42,18 +41,6 @@ func builtinTools() []ai.Tool {
 				},
 			},
 		},
-		{
-			Name:         toolHumanInput,
-			OriginalName: toolHumanInput,
-			Description: "Pause this agent run when you need missing information, a decision, or other human input before you can continue. " +
-				"The run is checkpointed as input-required and can be resumed with the human response without losing completed tool history.",
-			Properties: map[string]any{
-				"prompt": map[string]any{
-					"type":        "string",
-					"description": "The specific question, decision, or instruction needed from the human operator.",
-				},
-			},
-		},
 		{
 			Name:         toolDelegate,
 			OriginalName: toolDelegate,
@@ -91,9 +78,6 @@ func Builtins(opts ...Option) (tools []ai.Tool, handle func(name string, input m
 		case toolPlan:
 			r := a.handlePlan(ai.ToolCall{Name: name, Input: input})
 			return r.Value, r.Content, true
-		case toolHumanInput:
-			r := a.handleHumanInput(ai.ToolCall{Name: name, Input: input})
-			return r.Value, r.Content, true
 		case toolDelegate:
 			r := a.handleDelegate(context.Background(), ai.ToolCall{Name: name, Input: input})
 			return r.Value, r.Content, true
@@ -162,9 +146,6 @@ func (a *agentImpl) baseHandler() ai.ToolHandler {
 				return ai.ToolResult{ID: call.ID, Value: out, Content: out}
 			}
 		}
-		if call.Name == toolHumanInput {
-			return a.handleHumanInput(call)
-		}
 		if call.Name == toolDelegate {
 			return a.handleDelegate(ctx, call)
 		}
@@ -225,11 +206,6 @@ type approvalPause struct {
 	Message string
 }
 
-type inputPause struct {
-	OriginalMessage string `json:"original_message"`
-	Prompt          string `json:"prompt"`
-}
-
 func (a *agentImpl) approveWrap(next ai.ToolHandler) ai.ToolHandler {
 	return func(ctx context.Context, call ai.ToolCall) ai.ToolResult {
 		if a.opts.Approve != nil {
@@ -257,17 +233,6 @@ func (a *agentImpl) handlePlan(call ai.ToolCall) ai.ToolResult {
 	return ai.ToolResult{ID: call.ID, Value: call.Input, Content: string(data)}
 }
 
-// handleHumanInput records that the model needs operator input before it can continue.
-func (a *agentImpl) handleHumanInput(call ai.ToolCall) ai.ToolResult {
-	prompt, _ := call.Input["prompt"].(string)
-	prompt = strings.TrimSpace(prompt)
-	if prompt == "" {
-		prompt = "human input required"
-	}
-	a.pause = &approvalPause{Tool: toolHumanInput, Message: prompt}
-	return refused(call.ID, ai.RefusedApproval, "input-required: "+prompt)
-}
-
 // handleDelegate hands a subtask to another agent. Delegate-first:
 // if 'to' names a registered agent, it is called via RPC. Otherwise an
 // ephemeral sub-agent is created with a fresh, isolated context, asked

agent/builtin_test.go

@@ -11,15 +11,15 @@ import (
 
 func TestBuiltinTools(t *testing.T) {
 	tools := builtinTools()
-	if len(tools) != 3 {
-		t.Fatalf("builtinTools() = %d tools, want 3", len(tools))
+	if len(tools) != 2 {
+		t.Fatalf("builtinTools() = %d tools, want 2", len(tools))
 	}
 	names := map[string]bool{}
 	for _, tl := range tools {
 		names[tl.Name] = true
 	}
-	if !names[toolPlan] || !names[toolDelegate] || !names[toolHumanInput] {
-		t.Errorf("builtin tools = %v, want plan, request_input, and delegate", names)
+	if !names[toolPlan] || !names[toolDelegate] {
+		t.Errorf("builtin tools = %v, want plan and delegate", names)
 	}
 }
 
@@ -109,8 +109,8 @@ func TestBuiltinsAccessor(t *testing.T) {
 		WithRegistry(registry.NewMemoryRegistry()),
 	)
 
-	if len(tools) != 3 {
-		t.Fatalf("Builtins() returned %d tools, want 3", len(tools))
+	if len(tools) != 2 {
+		t.Fatalf("Builtins() returned %d tools, want 2", len(tools))
 	}
 
 	// A name that isn't a built-in falls through (ok == false).

agent/checkpoint.go

@@ -13,7 +13,6 @@ import (
 const (
 	agentAskStep      = "ask"
 	agentApprovalStep = "approval"
-	agentInputStep    = "input-required"
 )
 
 func (a *agentImpl) newCheckpointRun(runID, message, parentRunID string, existing *flow.Run) flow.Run {
@@ -52,17 +51,6 @@ func (a *agentImpl) saveRun(ctx context.Context, run flow.Run) error {
 	return nil
 }
 
-// Resume returns the response for a checkpointed agent run. Completed runs are
-// returned from the checkpoint without calling the model or replaying tool
-// calls; failed or in-progress runs continue from the saved input message.
-func Resume(ctx context.Context, ag Agent, runID string) (*Response, error) {
-	a, ok := ag.(*agentImpl)
-	if !ok {
-		return nil, fmt.Errorf("agent resume: unsupported agent implementation %T", ag)
-	}
-	return a.resume(ctx, runID)
-}
-
 func (a *agentImpl) resume(ctx context.Context, runID string) (*Response, error) {
 	if a.opts.Checkpoint == nil {
 		return nil, fmt.Errorf("agent %s has no checkpoint configured", a.opts.Name)
@@ -75,9 +63,6 @@ func (a *agentImpl) resume(ctx context.Context, runID string) (*Response, error)
 		return nil, fmt.Errorf("agent run %s not found", runID)
 	}
 	if run.Status == "paused" {
-		if run.State.Stage == agentInputStep {
-			return nil, fmt.Errorf("agent run %s is input-required; resume with ResumeInput", runID)
-		}
 		run.Status = "running"
 		run.State.Stage = agentAskStep
 	}
@@ -98,51 +83,6 @@ func (a *agentImpl) resume(ctx context.Context, runID string) (*Response, error)
 	return a.askLocked(ctx, run.ID, message, parentID, &run)
 }
 
-// ResumeInput resumes a checkpointed agent run that paused via the built-in
-// request_input tool. The supplied input is appended to the original request so
-// the same run can continue with durable checkpoint and completed tool history.
-func ResumeInput(ctx context.Context, ag Agent, runID, input string) (*Response, error) {
-	a, ok := ag.(*agentImpl)
-	if !ok {
-		return nil, fmt.Errorf("agent resume input: unsupported agent implementation %T", ag)
-	}
-	return a.resumeInput(ctx, runID, input)
-}
-
-func (a *agentImpl) resumeInput(ctx context.Context, runID, input string) (*Response, error) {
-	if a.opts.Checkpoint == nil {
-		return nil, fmt.Errorf("agent %s has no checkpoint configured", a.opts.Name)
-	}
-	run, ok, err := a.opts.Checkpoint.Load(ctx, runID)
-	if err != nil {
-		return nil, err
-	}
-	if !ok {
-		return nil, fmt.Errorf("agent run %s not found", runID)
-	}
-	if run.Status != "paused" || run.State.Stage != agentInputStep {
-		return nil, fmt.Errorf("agent run %s is not waiting for human input", runID)
-	}
-	var p inputPause
-	if err := run.State.Scan(&p); err != nil {
-		return nil, fmt.Errorf("agent run %s input state decode: %w", runID, err)
-	}
-	message := p.OriginalMessage
-	if message == "" {
-		message = string(run.State.Data)
-	}
-	message += "\n\nHuman input: " + input
-	run.Status = "running"
-	run.State.Stage = agentAskStep
-	run.State.Data = []byte(message)
-	a.mu.Lock()
-	defer a.mu.Unlock()
-	if a.model == nil {
-		a.setup()
-	}
-	return a.askLocked(ctx, run.ID, message, run.ParentID, &run)
-}
-
 func (a *agentImpl) pending(ctx context.Context) ([]flow.Run, error) {
 	if a.opts.Checkpoint == nil {
 		return nil, nil

agent/checkpoint_test.go

@@ -3,7 +3,6 @@ package agent
 import (
 	"context"
 	"errors"
-	"strings"
 	"testing"
 
 	"go-micro.dev/v6/ai"
@@ -119,64 +118,6 @@ func TestPendingReturnsUnfinishedAgentRuns(t *testing.T) {
 	}
 }
 
-func TestHumanInputPauseResumesSameRunWithInput(t *testing.T) {
-	ctx := context.Background()
-	cp := flow.StoreCheckpoint(store.NewStore(), "input-agent")
-	calls := 0
-	fakeGen = func(ctx context.Context, opts ai.Options, req *ai.Request) (*ai.Response, error) {
-		calls++
-		if calls == 1 {
-			if opts.ToolHandler != nil {
-				opts.ToolHandler(ctx, ai.ToolCall{ID: "input-1", Name: toolHumanInput, Input: map[string]any{"prompt": "Which region should I deploy to?"}})
-			}
-			return &ai.Response{Reply: "waiting"}, nil
-		}
-		if !strings.Contains(req.Prompt, "Human input: us-east-1") {
-			t.Fatalf("resumed prompt = %q, want human input", req.Prompt)
-		}
-		return &ai.Response{Reply: "deploying to us-east-1"}, nil
-	}
-	defer func() { fakeGen = nil }()
-
-	a := newTestAgent(Name("input-agent"), WithCheckpoint(cp))
-	_, err := a.Ask(ctx, "deploy the service")
-	if err == nil {
-		t.Fatal("Ask succeeded, want input-required pause")
-	}
-	runs, err := Pending(ctx, a)
-	if err != nil {
-		t.Fatalf("Pending: %v", err)
-	}
-	if len(runs) != 1 || runs[0].Status != "paused" || runs[0].State.Stage != agentInputStep {
-		t.Fatalf("paused runs = %#v, want one input-required run", runs)
-	}
-	var pause inputPause
-	if err := runs[0].State.Scan(&pause); err != nil {
-		t.Fatalf("Scan pause: %v", err)
-	}
-	if pause.OriginalMessage != "deploy the service" || pause.Prompt != "Which region should I deploy to?" {
-		t.Fatalf("pause = %#v", pause)
-	}
-
-	if _, err := Resume(ctx, a, runs[0].ID); err == nil || !strings.Contains(err.Error(), "ResumeInput") {
-		t.Fatalf("Resume input-required err = %v, want guidance", err)
-	}
-	resp, err := ResumeInput(ctx, a, runs[0].ID, "us-east-1")
-	if err != nil {
-		t.Fatalf("ResumeInput: %v", err)
-	}
-	if resp.RunID != runs[0].ID || resp.Reply != "deploying to us-east-1" {
-		t.Fatalf("response = %#v", resp)
-	}
-	loaded, ok, err := cp.Load(ctx, runs[0].ID)
-	if err != nil || !ok {
-		t.Fatalf("Load resumed run ok=%v err=%v", ok, err)
-	}
-	if loaded.Status != "done" {
-		t.Fatalf("resumed run status = %q, want done", loaded.Status)
-	}
-}
-
 func TestApprovalDenialPausesCheckpointedRunAndResumeContinues(t *testing.T) {
 	ctx := context.Background()
 	cp := flow.StoreCheckpoint(store.NewStore(), "approval-agent")

agent/memory.go

@@ -3,7 +3,6 @@ package agent
 import (
 	"encoding/json"
 	"fmt"
-	"sort"
 	"strings"
 	"sync"
 
@@ -124,31 +123,17 @@ func (m *storeMemory) Recall(query string, limit int) []ai.Message {
 		limit = 5
 	}
 	terms := recallTerms(query)
-	type match struct {
-		msg   ai.Message
-		score int
-		index int
-	}
-	matches := make([]match, 0, len(m.archive))
-	for i := len(m.archive) - 1; i >= 0; i-- {
+	var out []ai.Message
+	for i := len(m.archive) - 1; i >= 0 && len(out) < limit; i-- {
 		msg := m.archive[i]
-		if score := recallScore(msg, terms); score > 0 {
-			matches = append(matches, match{msg: msg, score: score, index: i})
+		text := strings.ToLower(fmt.Sprint(msg.Content))
+		for _, term := range terms {
+			if strings.Contains(text, term) {
+				out = append(out, msg)
+				break
+			}
 		}
 	}
-	sort.SliceStable(matches, func(i, j int) bool {
-		if matches[i].score != matches[j].score {
-			return matches[i].score > matches[j].score
-		}
-		return matches[i].index > matches[j].index
-	})
-	if len(matches) > limit {
-		matches = matches[:limit]
-	}
-	out := make([]ai.Message, 0, len(matches))
-	for _, match := range matches {
-		out = append(out, match.msg)
-	}
 	return out
 }
 
@@ -243,17 +228,6 @@ func compactText(s string, max int) string {
 	return s
 }
 
-func recallScore(msg ai.Message, terms []string) int {
-	text := strings.ToLower(fmt.Sprint(msg.Content))
-	score := 0
-	for _, term := range terms {
-		if strings.Contains(text, term) {
-			score++
-		}
-	}
-	return score
-}
-
 func recallTerms(query string) []string {
 	seen := map[string]bool{}
 	var terms []string

agent/memory_test.go

@@ -62,46 +62,6 @@ func TestWithMemoryUsed(t *testing.T) {
 	}
 }
 
-func TestCompactingMemoryRecallRanksSpecificMatches(t *testing.T) {
-	m := NewCompactingMemory(store.NewMemoryStore(), "agent/rank/history", 3, 1).(MemoryRecall)
-	writer := m.(Memory)
-	writer.Add("user", "alpha budget is 42")
-	writer.Add("assistant", "noted")
-	writer.Add("user", "beta budget is 7")
-	writer.Add("assistant", "noted")
-	writer.Add("user", "alpha owner is sam")
-
-	recalled := m.Recall("alpha budget", 2)
-	if len(recalled) == 0 {
-		t.Fatal("expected recalled messages")
-	}
-	if got := recalled[0].Content.(string); !strings.Contains(got, "alpha budget is 42") {
-		t.Fatalf("top recall = %q, want alpha budget match", got)
-	}
-}
-
-func TestCompactingMemoryArchivePersistsAndReloads(t *testing.T) {
-	st := store.NewMemoryStore()
-	m := NewCompactingMemory(st, "agent/reload/history", 3, 1)
-	m.Add("user", "alpha budget is 42")
-	m.Add("assistant", "noted")
-	m.Add("user", "beta budget is 7")
-	m.Add("assistant", "noted")
-
-	reloaded := NewCompactingMemory(st, "agent/reload/history", 3, 1)
-	recall, ok := reloaded.(MemoryRecall)
-	if !ok {
-		t.Fatal("compacting memory should support recall")
-	}
-	recalled := recall.Recall("alpha budget", 1)
-	if len(recalled) != 1 {
-		t.Fatalf("recalled %d messages, want 1", len(recalled))
-	}
-	if got := recalled[0].Content.(string); !strings.Contains(got, "alpha budget is 42") {
-		t.Fatalf("reloaded recall = %q, want alpha budget", got)
-	}
-}
-
 // A custom tool is offered to the model and dispatched to its handler.
 func TestWithToolExposedAndDispatched(t *testing.T) {
 	var got map[string]any

agent/otel.go

@@ -31,8 +31,6 @@ const (
 	AttrInputTokens    = "agent.tokens.input"
 	AttrOutputTokens   = "agent.tokens.output"
 	AttrTotalTokens    = "agent.tokens.total"
-	AttrAttempt        = "agent.model.attempt"
-	AttrMaxAttempts    = "agent.model.max_attempts"
 	AttrToolName       = "agent.tool.name"
 	AttrDelegate       = "agent.delegate"
 	AttrGuardrailBlock = "agent.guardrail.block"
@@ -40,22 +38,20 @@ const (
 )
 
 type RunEvent struct {
-	Time        time.Time `json:"time"`
-	RunID       string    `json:"run_id"`
-	ParentID    string    `json:"parent_id,omitempty"`
-	TraceID     string    `json:"trace_id,omitempty"`
-	SpanID      string    `json:"span_id,omitempty"`
-	Agent       string    `json:"agent"`
-	Kind        string    `json:"kind"`
-	Name        string    `json:"name,omitempty"`
-	Provider    string    `json:"provider,omitempty"`
-	Model       string    `json:"model,omitempty"`
-	Attempt     int       `json:"attempt,omitempty"`
-	MaxAttempts int       `json:"max_attempts,omitempty"`
-	LatencyMS   int64     `json:"latency_ms,omitempty"`
-	Tokens      Usage     `json:"tokens,omitempty"`
-	Refused     string    `json:"refused,omitempty"`
-	Error       string    `json:"error,omitempty"`
+	Time      time.Time `json:"time"`
+	RunID     string    `json:"run_id"`
+	ParentID  string    `json:"parent_id,omitempty"`
+	TraceID   string    `json:"trace_id,omitempty"`
+	SpanID    string    `json:"span_id,omitempty"`
+	Agent     string    `json:"agent"`
+	Kind      string    `json:"kind"`
+	Name      string    `json:"name,omitempty"`
+	Provider  string    `json:"provider,omitempty"`
+	Model     string    `json:"model,omitempty"`
+	LatencyMS int64     `json:"latency_ms,omitempty"`
+	Tokens    Usage     `json:"tokens,omitempty"`
+	Refused   string    `json:"refused,omitempty"`
+	Error     string    `json:"error,omitempty"`
 }
 
 type Usage = ai.Usage
@@ -148,7 +144,7 @@ func (m *tracedModel) Generate(ctx context.Context, req *ai.Request, opts ...ai.
 		if resp != nil {
 			usage = resp.Usage
 		}
-		e := RunEvent{Time: time.Now(), RunID: info.RunID, ParentID: info.ParentID, Agent: info.Agent, Kind: "model", Provider: provider, Model: model, Attempt: info.Attempt, MaxAttempts: info.MaxAttempts, LatencyMS: dur, Tokens: usage}
+		e := RunEvent{Time: time.Now(), RunID: info.RunID, ParentID: info.ParentID, Agent: info.Agent, Kind: "model", Provider: provider, Model: model, LatencyMS: dur, Tokens: usage}
 		if err != nil {
 			e.Error = err.Error()
 		}
@@ -166,12 +162,6 @@ func (m *tracedModel) Generate(ctx context.Context, req *ai.Request, opts ...ai.
 	resp, err := m.Model.Generate(ctx, req, opts...)
 	dur := time.Since(start).Milliseconds()
 	attrs := []attribute.KeyValue{attribute.Int64(AttrLatencyMS, dur)}
-	if info.Attempt > 0 {
-		attrs = append(attrs, attribute.Int(AttrAttempt, info.Attempt))
-	}
-	if info.MaxAttempts > 0 {
-		attrs = append(attrs, attribute.Int(AttrMaxAttempts, info.MaxAttempts))
-	}
 	usage := ai.Usage{}
 	if resp != nil {
 		usage = resp.Usage
@@ -185,7 +175,7 @@ func (m *tracedModel) Generate(ctx context.Context, req *ai.Request, opts ...ai.
 		span.SetStatus(codes.Ok, "")
 	}
 	span.End()
-	e := RunEvent{Time: time.Now(), RunID: info.RunID, ParentID: info.ParentID, Agent: info.Agent, Kind: "model", Provider: provider, Model: model, Attempt: info.Attempt, MaxAttempts: info.MaxAttempts, LatencyMS: dur, Tokens: usage}
+	e := RunEvent{Time: time.Now(), RunID: info.RunID, ParentID: info.ParentID, Agent: info.Agent, Kind: "model", Provider: provider, Model: model, LatencyMS: dur, Tokens: usage}
 	if err != nil {
 		e.Error = err.Error()
 	}
@@ -284,12 +274,6 @@ func runEventAttributes(e RunEvent) []attribute.KeyValue {
 	if e.Model != "" {
 		attrs = append(attrs, attribute.String(AttrModel, e.Model))
 	}
-	if e.Attempt > 0 {
-		attrs = append(attrs, attribute.Int(AttrAttempt, e.Attempt))
-	}
-	if e.MaxAttempts > 0 {
-		attrs = append(attrs, attribute.Int(AttrMaxAttempts, e.MaxAttempts))
-	}
 	if e.LatencyMS > 0 {
 		attrs = append(attrs, attribute.Int64(AttrLatencyMS, e.LatencyMS))
 	}

agent/otel_test.go

@@ -3,7 +3,6 @@ package agent
 import (
 	"context"
 	"encoding/json"
-	"errors"
 	"fmt"
 	"strings"
 	"testing"
@@ -12,13 +11,10 @@ import (
 	"go-micro.dev/v6/ai"
 	"go-micro.dev/v6/store"
 	"go.opentelemetry.io/otel/attribute"
-	"go.opentelemetry.io/otel/codes"
 	"go.opentelemetry.io/otel/sdk/trace"
 	"go.opentelemetry.io/otel/sdk/trace/tracetest"
 )
 
-const codesError = codes.Error
-
 type otelTestModel struct{ opts ai.Options }
 
 func (m *otelTestModel) Init(opts ...ai.Option) error {
@@ -93,9 +89,6 @@ func TestAgentOpenTelemetrySpans(t *testing.T) {
 		if attrs[AttrRunID] != runID || attrs[AttrAgentName] != "runner" {
 			t.Fatalf("%s missing run correlation attributes: %#v", s.Name(), attrs)
 		}
-		if s.Name() == spanNameModelCall && (attrs[AttrAttempt] != "1" || attrs[AttrMaxAttempts] != "1") {
-			t.Fatalf("model span missing attempt attributes: %#v", attrs)
-		}
 	}
 	keys, err := store.Scope(st, "agent", "runner").List(store.ListPrefix("runs/"))
 	if err != nil {
@@ -132,77 +125,6 @@ func TestAgentOpenTelemetrySpans(t *testing.T) {
 	}
 }
 
-type failingOtelModel struct{ opts ai.Options }
-
-func (m *failingOtelModel) Init(opts ...ai.Option) error {
-	for _, o := range opts {
-		o(&m.opts)
-	}
-	return nil
-}
-func (m *failingOtelModel) Options() ai.Options { return m.opts }
-func (m *failingOtelModel) String() string      { return "otelfail" }
-func (m *failingOtelModel) Stream(context.Context, *ai.Request, ...ai.GenerateOption) (ai.Stream, error) {
-	return nil, nil
-}
-func (m *failingOtelModel) Generate(context.Context, *ai.Request, ...ai.GenerateOption) (*ai.Response, error) {
-	return nil, errors.New("provider exploded")
-}
-
-func init() {
-	ai.Register("otelfail", func(opts ...ai.Option) ai.Model { return &failingOtelModel{opts: ai.NewOptions(opts...)} })
-}
-
-func TestAgentOpenTelemetrySpansModelFailure(t *testing.T) {
-	exp := tracetest.NewInMemoryExporter()
-	tp := trace.NewTracerProvider(trace.WithSyncer(exp))
-	st := store.NewMemoryStore()
-	a := New(Name("failing-runner"), Provider("otelfail"), WithStore(st), TraceProvider(tp))
-	if _, err := a.Ask(context.Background(), "hello"); err == nil {
-		t.Fatal("Ask succeeded, want provider error")
-	}
-
-	spans := exp.GetSpans().Snapshots()
-	var sawRunError, sawModelError bool
-	for _, s := range spans {
-		attrs := spanAttributes(s.Attributes())
-		switch s.Name() {
-		case spanNameRun:
-			if attrs[AttrAgentName] == "failing-runner" && s.Status().Code == codesError {
-				sawRunError = true
-			}
-		case spanNameModelCall:
-			if attrs[AttrAgentName] == "failing-runner" && attrs[AttrAttempt] == "1" && s.Status().Code == codesError {
-				sawModelError = true
-			}
-		}
-	}
-	if !sawRunError || !sawModelError {
-		t.Fatalf("missing error spans: run=%v model=%v spans=%d", sawRunError, sawModelError, len(spans))
-	}
-
-	summaries, err := ListRunSummaries(st, "failing-runner")
-	if err != nil {
-		t.Fatal(err)
-	}
-	if len(summaries) != 1 || summaries[0].Status != "error" || summaries[0].LastError == "" {
-		t.Fatalf("unexpected failure summary: %#v", summaries)
-	}
-	events, err := LoadRunEvents(st, "failing-runner", summaries[0].RunID)
-	if err != nil {
-		t.Fatal(err)
-	}
-	var sawModelEvent bool
-	for _, event := range events {
-		if event.Kind == "model" && event.Attempt == 1 && event.MaxAttempts == 1 && event.Error != "" {
-			sawModelEvent = true
-		}
-	}
-	if !sawModelEvent {
-		t.Fatalf("missing failed model event with attempt metadata: %#v", events)
-	}
-}
-
 func spanEventHasRunInfo(events []trace.Event, name, runID, agentName string) bool {
 	for _, event := range events {
 		if event.Name != name {

ai/anthropic/anthropic.go

@@ -77,9 +77,11 @@ func (p *Provider) Generate(ctx context.Context, req *ai.Request, opts ...ai.Gen
 	// Build initial request
 	apiReq := map[string]any{
 		"model":      p.opts.Model,
-		"max_tokens": anthropicMaxTokens(p.opts),
+		"max_tokens": 8192,
 		"system":     req.SystemPrompt,
-		"messages":   threadAnthropicMessages(req),
+		"messages": []map[string]any{
+			{"role": "user", "content": req.Prompt},
+		},
 	}
 
 	if len(anthropicTools) > 0 {
@@ -99,9 +101,10 @@ func (p *Provider) Generate(ctx context.Context, req *ai.Request, opts ...ai.Gen
 
 	// Tool execution loop: execute tools, send results back, repeat
 	// until the model responds with text only (no more tool calls)
-	messages := append(threadAnthropicMessages(req),
-		map[string]any{"role": "assistant", "content": cleanContent(rawContent)},
-	)
+	messages := []map[string]any{
+		{"role": "user", "content": req.Prompt},
+		{"role": "assistant", "content": cleanContent(rawContent)},
+	}
 
 	pendingCalls := resp.ToolCalls
 
@@ -124,7 +127,7 @@ func (p *Provider) Generate(ctx context.Context, req *ai.Request, opts ...ai.Gen
 
 		followUpReq := map[string]any{
 			"model":      p.opts.Model,
-			"max_tokens": anthropicMaxTokens(p.opts),
+			"max_tokens": 8192,
 			"system":     req.SystemPrompt,
 			"messages":   messages,
 		}
@@ -267,24 +270,3 @@ func cleanContent(raw any) any {
 	}
 	return cleaned
 }
-
-// threadAnthropicMessages builds the Anthropic messages array from the
-// conversation history (req.Messages) followed by the current prompt. The
-// system prompt is sent separately via the top-level "system" field.
-func threadAnthropicMessages(req *ai.Request) []map[string]any {
-	msgs := make([]map[string]any, 0, len(req.Messages)+1)
-	for _, m := range req.Messages {
-		msgs = append(msgs, map[string]any{"role": m.Role, "content": m.Content})
-	}
-	if req.Prompt != "" {
-		msgs = append(msgs, map[string]any{"role": "user", "content": req.Prompt})
-	}
-	return msgs
-}
-
-func anthropicMaxTokens(o ai.Options) int {
-	if o.MaxTokens > 0 {
-		return o.MaxTokens
-	}
-	return 8192
-}

ai/atlascloud/atlascloud.go

@@ -20,7 +20,6 @@
 package atlascloud
 
 import (
-	"bufio"
 	"bytes"
 	"context"
 	"encoding/json"
@@ -43,7 +42,6 @@ func init() {
 	ai.RegisterVideo("atlascloud", func(opts ...ai.Option) ai.VideoModel {
 		return NewProvider(opts...)
 	})
-	ai.RegisterStream("atlascloud")
 }
 
 // Provider implements the ai.Model interface for Atlas Cloud.
@@ -93,21 +91,13 @@ func (p *Provider) Generate(ctx context.Context, req *ai.Request, opts ...ai.Gen
 
 	messages := []map[string]any{
 		{"role": "system", "content": req.SystemPrompt},
-	}
-	for _, m := range req.Messages {
-		messages = append(messages, map[string]any{"role": m.Role, "content": m.Content})
-	}
-	if req.Prompt != "" {
-		messages = append(messages, map[string]any{"role": "user", "content": req.Prompt})
+		{"role": "user", "content": req.Prompt},
 	}
 
 	apiReq := map[string]any{
 		"model":    p.opts.Model,
 		"messages": messages,
 	}
-	if p.opts.MaxTokens > 0 {
-		apiReq["max_tokens"] = p.opts.MaxTokens
-	}
 
 	if len(tools) > 0 {
 		apiReq["tools"] = tools
@@ -152,111 +142,8 @@ func (p *Provider) Generate(ctx context.Context, req *ai.Request, opts ...ai.Gen
 	return resp, nil
 }
 
-// Stream generates a streaming response from Atlas Cloud's OpenAI-compatible
-// chat completions endpoint, emitting content deltas as they arrive.
 func (p *Provider) Stream(ctx context.Context, req *ai.Request, opts ...ai.GenerateOption) (ai.Stream, error) {
-	messages := []map[string]any{
-		{"role": "system", "content": req.SystemPrompt},
-	}
-	for _, m := range req.Messages {
-		messages = append(messages, map[string]any{"role": m.Role, "content": m.Content})
-	}
-	if req.Prompt != "" {
-		messages = append(messages, map[string]any{"role": "user", "content": req.Prompt})
-	}
-	apiReq := map[string]any{
-		"model":          p.opts.Model,
-		"messages":       messages,
-		"stream":         true,
-		"stream_options": map[string]any{"include_usage": true},
-	}
-	if p.opts.MaxTokens > 0 {
-		apiReq["max_tokens"] = p.opts.MaxTokens
-	}
-	reqBody, err := json.Marshal(apiReq)
-	if err != nil {
-		return nil, fmt.Errorf("failed to marshal stream request: %w", err)
-	}
-	apiURL := strings.TrimRight(p.opts.BaseURL, "/") + "/v1/chat/completions"
-	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, apiURL, bytes.NewReader(reqBody))
-	if err != nil {
-		return nil, fmt.Errorf("failed to create stream request: %w", err)
-	}
-	httpReq.Header.Set("Content-Type", "application/json")
-	httpReq.Header.Set("Accept", "text/event-stream")
-	httpReq.Header.Set("Authorization", "Bearer "+p.opts.APIKey)
-
-	httpResp, err := http.DefaultClient.Do(httpReq)
-	if err != nil {
-		return nil, fmt.Errorf("stream API request failed: %w", err)
-	}
-	if httpResp.StatusCode != http.StatusOK {
-		defer httpResp.Body.Close()
-		respBody, _ := io.ReadAll(httpResp.Body)
-		return nil, fmt.Errorf("stream API error (%s): %s", httpResp.Status, string(respBody))
-	}
-	return &atlasStream{body: httpResp.Body, scanner: bufio.NewScanner(httpResp.Body)}, nil
-}
-
-type atlasStream struct {
-	body    io.ReadCloser
-	scanner *bufio.Scanner
-	closed  bool
-}
-
-func (s *atlasStream) Recv() (*ai.Response, error) {
-	for s.scanner.Scan() {
-		line := strings.TrimSpace(s.scanner.Text())
-		if line == "" || strings.HasPrefix(line, ":") {
-			continue
-		}
-		if !strings.HasPrefix(line, "data:") {
-			continue
-		}
-		data := strings.TrimSpace(strings.TrimPrefix(line, "data:"))
-		if data == "[DONE]" {
-			return nil, io.EOF
-		}
-		var chunk struct {
-			Choices []struct {
-				Delta struct {
-					Content string `json:"content"`
-				} `json:"delta"`
-			} `json:"choices"`
-			Usage *struct {
-				PromptTokens     int `json:"prompt_tokens"`
-				CompletionTokens int `json:"completion_tokens"`
-				TotalTokens      int `json:"total_tokens"`
-			} `json:"usage"`
-		}
-		if err := json.Unmarshal([]byte(data), &chunk); err != nil {
-			return nil, fmt.Errorf("failed to parse stream chunk: %w", err)
-		}
-		if len(chunk.Choices) > 0 && chunk.Choices[0].Delta.Content != "" {
-			return &ai.Response{Reply: chunk.Choices[0].Delta.Content}, nil
-		}
-		// Final chunk (after include_usage) carries token usage and no content.
-		if chunk.Usage != nil {
-			return &ai.Response{Usage: ai.Usage{
-				InputTokens:  chunk.Usage.PromptTokens,
-				OutputTokens: chunk.Usage.CompletionTokens,
-				TotalTokens:  chunk.Usage.TotalTokens,
-			}}, nil
-		}
-		continue
-	}
-	if err := s.scanner.Err(); err != nil {
-		return nil, err
-	}
-	return nil, io.EOF
-}
-
-func (s *atlasStream) Close() error {
-	if s.closed {
-		return nil
-	}
-	s.closed = true
-	return s.body.Close()
+	return nil, fmt.Errorf("%w: atlascloud provider", ai.ErrStreamingUnsupported)
 }
 
 func (p *Provider) callAPI(ctx context.Context, req map[string]any) (*ai.Response, map[string]any, error) {

ai/atlascloud/atlascloud_test.go

@@ -2,11 +2,7 @@ package atlascloud
 
 import (
 	"context"
-	"encoding/json"
 	"errors"
-	"io"
-	"net/http"
-	"net/http/httptest"
 	"testing"
 
 	"go-micro.dev/v6/ai"
@@ -85,58 +81,16 @@ func TestProvider_Generate_NoAPIKey(t *testing.T) {
 	}
 }
 
-func TestProvider_Stream(t *testing.T) {
-	var sawStream, sawIncludeUsage bool
-	ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		if r.URL.Path != "/v1/chat/completions" {
-			t.Errorf("path = %s, want /v1/chat/completions", r.URL.Path)
-		}
-		var body map[string]any
-		if err := json.NewDecoder(r.Body).Decode(&body); err != nil {
-			t.Fatalf("decode request: %v", err)
-		}
-		sawStream, _ = body["stream"].(bool)
-		if so, ok := body["stream_options"].(map[string]any); ok {
-			sawIncludeUsage, _ = so["include_usage"].(bool)
-		}
-		w.Header().Set("Content-Type", "text/event-stream")
-		_, _ = w.Write([]byte("data: {\"choices\":[{\"delta\":{\"content\":\"hel\"}}]}\n\n"))
-		_, _ = w.Write([]byte("data: {\"choices\":[{\"delta\":{\"content\":\"lo\"}}]}\n\n"))
-		_, _ = w.Write([]byte("data: {\"choices\":[],\"usage\":{\"prompt_tokens\":7,\"completion_tokens\":2,\"total_tokens\":9}}\n\n"))
-		_, _ = w.Write([]byte("data: [DONE]\n\n"))
-	}))
-	defer ts.Close()
+func TestProvider_Stream_NotImplemented(t *testing.T) {
+	p := NewProvider()
 
-	p := NewProvider(ai.WithAPIKey("test-key"), ai.WithBaseURL(ts.URL))
-	stream, err := p.Stream(context.Background(), &ai.Request{Prompt: "Hello"})
-	if err != nil {
-		t.Fatalf("Stream returned error: %v", err)
-	}
-	defer stream.Close()
-	if !sawStream {
-		t.Fatal("stream request did not set stream=true")
-	}
-	if !sawIncludeUsage {
-		t.Fatal("stream request did not set stream_options.include_usage=true")
+	req := &ai.Request{
+		Prompt: "Hello",
 	}
 
-	first, err := stream.Recv()
-	if err != nil || first.Reply != "hel" {
-		t.Fatalf("first chunk = %#v, %v; want hel", first, err)
-	}
-	second, err := stream.Recv()
-	if err != nil || second.Reply != "lo" {
-		t.Fatalf("second chunk = %#v, %v; want lo", second, err)
-	}
-	usage, err := stream.Recv()
-	if err != nil {
-		t.Fatalf("usage chunk error: %v", err)
-	}
-	if usage.Usage.TotalTokens != 9 || usage.Usage.InputTokens != 7 || usage.Usage.OutputTokens != 2 {
-		t.Fatalf("usage = %#v; want input=7 output=2 total=9", usage.Usage)
-	}
-	if _, err := stream.Recv(); !errors.Is(err, io.EOF) {
-		t.Fatalf("final error = %v, want EOF", err)
+	_, err := p.Stream(context.Background(), req)
+	if !errors.Is(err, ai.ErrStreamingUnsupported) {
+		t.Fatalf("Stream error = %v, want ErrStreamingUnsupported", err)
 	}
 }
 

ai/capabilities_test.go

@@ -34,7 +34,7 @@ func TestRegisteredProviders(t *testing.T) {
 	}
 
 	got = ai.RegisteredProviders("stream")
-	want = []string{"atlascloud", "groq", "mistral", "openai", "together"}
+	want = []string{"openai"}
 	if !reflect.DeepEqual(got, want) {
 		t.Fatalf("RegisteredProviders(stream) = %#v, want %#v", got, want)
 	}
@@ -44,12 +44,12 @@ func TestCapabilityRows(t *testing.T) {
 	got := ai.CapabilityRows()
 	want := []ai.CapabilityRow{
 		{Provider: "anthropic", Capabilities: ai.Capabilities{Model: true}},
-		{Provider: "atlascloud", Capabilities: ai.Capabilities{Model: true, Image: true, Video: true, Stream: true}},
+		{Provider: "atlascloud", Capabilities: ai.Capabilities{Model: true, Image: true, Video: true}},
 		{Provider: "gemini", Capabilities: ai.Capabilities{Model: true}},
-		{Provider: "groq", Capabilities: ai.Capabilities{Model: true, Stream: true}},
-		{Provider: "mistral", Capabilities: ai.Capabilities{Model: true, Stream: true}},
+		{Provider: "groq", Capabilities: ai.Capabilities{Model: true}},
+		{Provider: "mistral", Capabilities: ai.Capabilities{Model: true}},
 		{Provider: "openai", Capabilities: ai.Capabilities{Model: true, Image: true, Stream: true}},
-		{Provider: "together", Capabilities: ai.Capabilities{Model: true, Stream: true}},
+		{Provider: "together", Capabilities: ai.Capabilities{Model: true}},
 	}
 	if !reflect.DeepEqual(got, want) {
 		t.Fatalf("CapabilityRows() = %#v, want %#v", got, want)
@@ -72,7 +72,7 @@ func TestCapabilityMatrix(t *testing.T) {
 	if caps := ai.ProviderCapabilities("openai"); caps != (ai.Capabilities{Model: true, Image: true, Stream: true}) {
 		t.Fatalf("ProviderCapabilities(openai) = %#v", caps)
 	}
-	if caps := ai.ProviderCapabilities("atlascloud"); caps != (ai.Capabilities{Model: true, Image: true, Video: true, Stream: true}) {
+	if caps := ai.ProviderCapabilities("atlascloud"); caps != (ai.Capabilities{Model: true, Image: true, Video: true}) {
 		t.Fatalf("ProviderCapabilities(atlascloud) = %#v", caps)
 	}
 	if caps := ai.ProviderCapabilities("missing"); caps != (ai.Capabilities{}) {
@@ -88,7 +88,7 @@ func TestRegisterStream(t *testing.T) {
 	}
 
 	got := ai.RegisteredProviders("stream")
-	want := []string{"atlascloud", "groq", "mistral", "openai", "test-stream", "together"}
+	want := []string{"openai", "test-stream"}
 	if !reflect.DeepEqual(got, want) {
 		t.Fatalf("RegisteredProviders(stream) = %#v, want %#v", got, want)
 	}

ai/groq/groq.go

@@ -22,14 +22,12 @@ import (
 	"strings"
 
 	"go-micro.dev/v6/ai"
-	"go-micro.dev/v6/ai/internal/openaiapi"
 )
 
 func init() {
 	ai.Register("groq", func(opts ...ai.Option) ai.Model {
 		return NewProvider(opts...)
 	})
-	ai.RegisterStream("groq")
 }
 
 type Provider struct {
@@ -121,7 +119,7 @@ func (p *Provider) Generate(ctx context.Context, req *ai.Request, opts ...ai.Gen
 }
 
 func (p *Provider) Stream(ctx context.Context, req *ai.Request, opts ...ai.GenerateOption) (ai.Stream, error) {
-	return openaiapi.Stream(ctx, p.opts, req, "/v1/chat/completions")
+	return nil, fmt.Errorf("%w: groq provider", ai.ErrStreamingUnsupported)
 }
 
 func (p *Provider) callAPI(ctx context.Context, req map[string]any) (*ai.Response, map[string]any, error) {

ai/groq/groq_test.go

@@ -2,11 +2,7 @@ package groq
 
 import (
 	"context"
-	"encoding/json"
 	"errors"
-	"io"
-	"net/http"
-	"net/http/httptest"
 	"testing"
 
 	"go-micro.dev/v6/ai"
@@ -44,44 +40,9 @@ func TestProvider_Generate_NoAPIKey(t *testing.T) {
 	}
 }
 
-func TestProvider_Stream(t *testing.T) {
-	var sawStream bool
-	ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		if r.URL.Path != "/v1/chat/completions" {
-			t.Fatalf("path = %s, want /v1/chat/completions", r.URL.Path)
-		}
-		var body map[string]any
-		if err := json.NewDecoder(r.Body).Decode(&body); err != nil {
-			t.Fatalf("decode request: %v", err)
-		}
-		sawStream, _ = body["stream"].(bool)
-		w.Header().Set("Content-Type", "text/event-stream")
-		_, _ = w.Write([]byte("data: {\"choices\":[{\"delta\":{\"content\":\"hel\"}}]}\n\n"))
-		_, _ = w.Write([]byte("data: {\"choices\":[{\"delta\":{\"content\":\"lo\"}}]}\n\n"))
-		_, _ = w.Write([]byte("data: [DONE]\n\n"))
-	}))
-	defer ts.Close()
-
-	p := NewProvider(ai.WithAPIKey("test-key"), ai.WithBaseURL(ts.URL))
-	stream, err := p.Stream(context.Background(), &ai.Request{Prompt: "Hello"})
-	if err != nil {
-		t.Fatalf("Stream returned error: %v", err)
-	}
-	defer stream.Close()
-	if !sawStream {
-		t.Fatal("stream request did not set stream=true")
-	}
-
-	first, err := stream.Recv()
-	if err != nil || first.Reply != "hel" {
-		t.Fatalf("first chunk = %#v, %v; want hel", first, err)
-	}
-	second, err := stream.Recv()
-	if err != nil || second.Reply != "lo" {
-		t.Fatalf("second chunk = %#v, %v; want lo", second, err)
-	}
-	if _, err := stream.Recv(); !errors.Is(err, io.EOF) {
-		t.Fatalf("final error = %v, want EOF", err)
+func TestProvider_Stream_NotImplemented(t *testing.T) {
+	if _, err := NewProvider().Stream(context.Background(), &ai.Request{Prompt: "hi"}); !errors.Is(err, ai.ErrStreamingUnsupported) {
+		t.Fatalf("Stream error = %v, want ErrStreamingUnsupported", err)
 	}
 }
 

ai/internal/openaiapi/stream.go

@@ -1,117 +0,0 @@
-package openaiapi
-
-import (
-	"bufio"
-	"bytes"
-	"context"
-	"encoding/json"
-	"fmt"
-	"io"
-	"net/http"
-	"strings"
-
-	"go-micro.dev/v6/ai"
-)
-
-// Stream opens an OpenAI-compatible chat completions SSE stream.
-func Stream(ctx context.Context, opts ai.Options, req *ai.Request, basePath string) (ai.Stream, error) {
-	messages := []map[string]any{{"role": "system", "content": req.SystemPrompt}}
-	for _, m := range req.Messages {
-		messages = append(messages, map[string]any{"role": m.Role, "content": m.Content})
-	}
-	if req.Prompt != "" {
-		messages = append(messages, map[string]any{"role": "user", "content": req.Prompt})
-	}
-	apiReq := map[string]any{
-		"model":          opts.Model,
-		"messages":       messages,
-		"stream":         true,
-		"stream_options": map[string]any{"include_usage": true},
-	}
-	if opts.MaxTokens > 0 {
-		apiReq["max_tokens"] = opts.MaxTokens
-	}
-	reqBody, err := json.Marshal(apiReq)
-	if err != nil {
-		return nil, fmt.Errorf("failed to marshal stream request: %w", err)
-	}
-	apiURL := strings.TrimRight(opts.BaseURL, "/") + basePath
-	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, apiURL, bytes.NewReader(reqBody))
-	if err != nil {
-		return nil, fmt.Errorf("failed to create stream request: %w", err)
-	}
-	httpReq.Header.Set("Content-Type", "application/json")
-	httpReq.Header.Set("Accept", "text/event-stream")
-	httpReq.Header.Set("Authorization", "Bearer "+opts.APIKey)
-
-	httpResp, err := http.DefaultClient.Do(httpReq)
-	if err != nil {
-		return nil, fmt.Errorf("stream API request failed: %w", err)
-	}
-	if httpResp.StatusCode != http.StatusOK {
-		defer httpResp.Body.Close()
-		respBody, _ := io.ReadAll(httpResp.Body)
-		return nil, fmt.Errorf("stream API error (%s): %s", httpResp.Status, string(respBody))
-	}
-	return &StreamReader{body: httpResp.Body, scanner: bufio.NewScanner(httpResp.Body)}, nil
-}
-
-// StreamReader reads OpenAI-compatible server-sent event chunks.
-type StreamReader struct {
-	body    io.ReadCloser
-	scanner *bufio.Scanner
-	closed  bool
-}
-
-func (s *StreamReader) Recv() (*ai.Response, error) {
-	for s.scanner.Scan() {
-		line := strings.TrimSpace(s.scanner.Text())
-		if line == "" || strings.HasPrefix(line, ":") {
-			continue
-		}
-		if !strings.HasPrefix(line, "data:") {
-			continue
-		}
-		data := strings.TrimSpace(strings.TrimPrefix(line, "data:"))
-		if data == "[DONE]" {
-			return nil, io.EOF
-		}
-		var chunk struct {
-			Choices []struct {
-				Delta struct {
-					Content string `json:"content"`
-				} `json:"delta"`
-			} `json:"choices"`
-			Usage *struct {
-				PromptTokens     int `json:"prompt_tokens"`
-				CompletionTokens int `json:"completion_tokens"`
-				TotalTokens      int `json:"total_tokens"`
-			} `json:"usage"`
-		}
-		if err := json.Unmarshal([]byte(data), &chunk); err != nil {
-			return nil, fmt.Errorf("failed to parse stream chunk: %w", err)
-		}
-		if len(chunk.Choices) > 0 && chunk.Choices[0].Delta.Content != "" {
-			return &ai.Response{Reply: chunk.Choices[0].Delta.Content}, nil
-		}
-		if chunk.Usage != nil {
-			return &ai.Response{Usage: ai.Usage{
-				InputTokens:  chunk.Usage.PromptTokens,
-				OutputTokens: chunk.Usage.CompletionTokens,
-				TotalTokens:  chunk.Usage.TotalTokens,
-			}}, nil
-		}
-	}
-	if err := s.scanner.Err(); err != nil {
-		return nil, err
-	}
-	return nil, io.EOF
-}
-
-func (s *StreamReader) Close() error {
-	if s.closed {
-		return nil
-	}
-	s.closed = true
-	return s.body.Close()
-}

ai/mistral/mistral.go

@@ -22,14 +22,12 @@ import (
 	"strings"
 
 	"go-micro.dev/v6/ai"
-	"go-micro.dev/v6/ai/internal/openaiapi"
 )
 
 func init() {
 	ai.Register("mistral", func(opts ...ai.Option) ai.Model {
 		return NewProvider(opts...)
 	})
-	ai.RegisterStream("mistral")
 }
 
 type Provider struct {
@@ -121,7 +119,7 @@ func (p *Provider) Generate(ctx context.Context, req *ai.Request, opts ...ai.Gen
 }
 
 func (p *Provider) Stream(ctx context.Context, req *ai.Request, opts ...ai.GenerateOption) (ai.Stream, error) {
-	return openaiapi.Stream(ctx, p.opts, req, "/v1/chat/completions")
+	return nil, fmt.Errorf("%w: mistral provider", ai.ErrStreamingUnsupported)
 }
 
 func (p *Provider) callAPI(ctx context.Context, req map[string]any) (*ai.Response, map[string]any, error) {

ai/mistral/mistral_test.go

@@ -2,11 +2,7 @@ package mistral
 
 import (
 	"context"
-	"encoding/json"
 	"errors"
-	"io"
-	"net/http"
-	"net/http/httptest"
 	"testing"
 
 	"go-micro.dev/v6/ai"
@@ -44,44 +40,9 @@ func TestProvider_Generate_NoAPIKey(t *testing.T) {
 	}
 }
 
-func TestProvider_Stream(t *testing.T) {
-	var sawStream bool
-	ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		if r.URL.Path != "/v1/chat/completions" {
-			t.Fatalf("path = %s, want /v1/chat/completions", r.URL.Path)
-		}
-		var body map[string]any
-		if err := json.NewDecoder(r.Body).Decode(&body); err != nil {
-			t.Fatalf("decode request: %v", err)
-		}
-		sawStream, _ = body["stream"].(bool)
-		w.Header().Set("Content-Type", "text/event-stream")
-		_, _ = w.Write([]byte("data: {\"choices\":[{\"delta\":{\"content\":\"hel\"}}]}\n\n"))
-		_, _ = w.Write([]byte("data: {\"choices\":[{\"delta\":{\"content\":\"lo\"}}]}\n\n"))
-		_, _ = w.Write([]byte("data: [DONE]\n\n"))
-	}))
-	defer ts.Close()
-
-	p := NewProvider(ai.WithAPIKey("test-key"), ai.WithBaseURL(ts.URL))
-	stream, err := p.Stream(context.Background(), &ai.Request{Prompt: "Hello"})
-	if err != nil {
-		t.Fatalf("Stream returned error: %v", err)
-	}
-	defer stream.Close()
-	if !sawStream {
-		t.Fatal("stream request did not set stream=true")
-	}
-
-	first, err := stream.Recv()
-	if err != nil || first.Reply != "hel" {
-		t.Fatalf("first chunk = %#v, %v; want hel", first, err)
-	}
-	second, err := stream.Recv()
-	if err != nil || second.Reply != "lo" {
-		t.Fatalf("second chunk = %#v, %v; want lo", second, err)
-	}
-	if _, err := stream.Recv(); !errors.Is(err, io.EOF) {
-		t.Fatalf("final error = %v, want EOF", err)
+func TestProvider_Stream_NotImplemented(t *testing.T) {
+	if _, err := NewProvider().Stream(context.Background(), &ai.Request{Prompt: "hi"}); !errors.Is(err, ai.ErrStreamingUnsupported) {
+		t.Fatalf("Stream error = %v, want ErrStreamingUnsupported", err)
 	}
 }
 

ai/openai/openai.go

@@ -85,12 +85,7 @@ func (p *Provider) Generate(ctx context.Context, req *ai.Request, opts ...ai.Gen
 	// Build messages
 	messages := []map[string]any{
 		{"role": "system", "content": req.SystemPrompt},
-	}
-	for _, m := range req.Messages {
-		messages = append(messages, map[string]any{"role": m.Role, "content": m.Content})
-	}
-	if req.Prompt != "" {
-		messages = append(messages, map[string]any{"role": "user", "content": req.Prompt})
+		{"role": "user", "content": req.Prompt},
 	}
 
 	// Build initial request
@@ -98,9 +93,6 @@ func (p *Provider) Generate(ctx context.Context, req *ai.Request, opts ...ai.Gen
 		"model":    p.opts.Model,
 		"messages": messages,
 	}
-	if p.opts.MaxTokens > 0 {
-		apiReq["max_tokens"] = p.opts.MaxTokens
-	}
 
 	if len(openaiTools) > 0 {
 		apiReq["tools"] = openaiTools
@@ -154,21 +146,12 @@ func (p *Provider) Generate(ctx context.Context, req *ai.Request, opts ...ai.Gen
 func (p *Provider) Stream(ctx context.Context, req *ai.Request, opts ...ai.GenerateOption) (ai.Stream, error) {
 	messages := []map[string]any{
 		{"role": "system", "content": req.SystemPrompt},
-	}
-	for _, m := range req.Messages {
-		messages = append(messages, map[string]any{"role": m.Role, "content": m.Content})
-	}
-	if req.Prompt != "" {
-		messages = append(messages, map[string]any{"role": "user", "content": req.Prompt})
+		{"role": "user", "content": req.Prompt},
 	}
 	apiReq := map[string]any{
-		"model":          p.opts.Model,
-		"messages":       messages,
-		"stream":         true,
-		"stream_options": map[string]any{"include_usage": true},
-	}
-	if p.opts.MaxTokens > 0 {
-		apiReq["max_tokens"] = p.opts.MaxTokens
+		"model":    p.opts.Model,
+		"messages": messages,
+		"stream":   true,
 	}
 	reqBody, err := json.Marshal(apiReq)
 	if err != nil {
@@ -220,27 +203,14 @@ func (s *openAIStream) Recv() (*ai.Response, error) {
 					Content string `json:"content"`
 				} `json:"delta"`
 			} `json:"choices"`
-			Usage *struct {
-				PromptTokens     int `json:"prompt_tokens"`
-				CompletionTokens int `json:"completion_tokens"`
-				TotalTokens      int `json:"total_tokens"`
-			} `json:"usage"`
 		}
 		if err := json.Unmarshal([]byte(data), &chunk); err != nil {
 			return nil, fmt.Errorf("failed to parse stream chunk: %w", err)
 		}
-		if len(chunk.Choices) > 0 && chunk.Choices[0].Delta.Content != "" {
-			return &ai.Response{Reply: chunk.Choices[0].Delta.Content}, nil
+		if len(chunk.Choices) == 0 || chunk.Choices[0].Delta.Content == "" {
+			continue
 		}
-		// Final chunk (after include_usage) carries token usage and no content.
-		if chunk.Usage != nil {
-			return &ai.Response{Usage: ai.Usage{
-				InputTokens:  chunk.Usage.PromptTokens,
-				OutputTokens: chunk.Usage.CompletionTokens,
-				TotalTokens:  chunk.Usage.TotalTokens,
-			}}, nil
-		}
-		continue
+		return &ai.Response{Reply: chunk.Choices[0].Delta.Content}, nil
 	}
 	if err := s.scanner.Err(); err != nil {
 		return nil, err

ai/options.go

@@ -16,8 +16,6 @@ type Options struct {
 	BaseURL string
 	// ToolHandler handles tool calls (optional, for automatic tool execution)
 	ToolHandler ToolHandler
-	// MaxTokens caps the length of the response (0 = provider default)
-	MaxTokens int
 }
 
 // GenerateOptions for generate call
@@ -93,11 +91,3 @@ func WithTools(t *Tools) Option {
 		}
 	}
 }
-
-// WithMaxTokens caps the number of tokens in the response. 0 leaves the
-// provider default in place.
-func WithMaxTokens(n int) Option {
-	return func(o *Options) {
-		o.MaxTokens = n
-	}
-}

ai/retry.go

@@ -13,12 +13,6 @@ type StatusCoder interface {
 	StatusCode() int
 }
 
-// RetryAfterCoder is implemented by provider errors that expose a server
-// supplied retry delay, such as HTTP Retry-After on a 429/503 response.
-type RetryAfterCoder interface {
-	RetryAfter() time.Duration
-}
-
 // ErrorKind classifies provider-boundary failures into stable buckets callers
 // can inspect without parsing provider-specific error strings.
 type ErrorKind string
@@ -119,7 +113,16 @@ func GenerateWithRetry(ctx context.Context, m Model, req *Request, policy Genera
 		// Always back off between retries — exponential and capped — so an
 		// opt-in retry can never become a tight loop hammering the provider,
 		// even if Backoff was left at zero.
-		backoff := retryBackoff(err, attempt, policy.Backoff)
+		backoff := policy.Backoff
+		if backoff <= 0 {
+			backoff = 200 * time.Millisecond
+		}
+		if shift := attempt - 1; shift > 0 {
+			backoff <<= shift
+		}
+		if backoff > 30*time.Second {
+			backoff = 30 * time.Second
+		}
 		t := time.NewTimer(backoff)
 		select {
 		case <-ctx.Done():
@@ -133,30 +136,6 @@ func GenerateWithRetry(ctx context.Context, m Model, req *Request, policy Genera
 	return nil, &RetryError{Attempts: policy.MaxAttempts, Kind: ClassifyError(last), Err: last}
 }
 
-func retryBackoff(err error, attempt int, base time.Duration) time.Duration {
-	backoff := base
-	if backoff <= 0 {
-		backoff = 200 * time.Millisecond
-	}
-	if shift := attempt - 1; shift > 0 {
-		backoff <<= shift
-	}
-	if backoff > 30*time.Second {
-		backoff = 30 * time.Second
-	}
-
-	var retryAfter RetryAfterCoder
-	if errors.As(err, &retryAfter) {
-		if delay := retryAfter.RetryAfter(); delay > backoff {
-			backoff = delay
-		}
-	}
-	if backoff > 30*time.Second {
-		return 30 * time.Second
-	}
-	return backoff
-}
-
 // ClassifyError maps provider and context failures to stable operational kinds.
 func ClassifyError(err error) ErrorKind {
 	if err == nil {

ai/retry_test.go

@@ -139,14 +139,6 @@ type statusErr int
 func (e statusErr) Error() string   { return "provider status" }
 func (e statusErr) StatusCode() int { return int(e) }
 
-type retryAfterErr struct {
-	delay time.Duration
-}
-
-func (e retryAfterErr) Error() string             { return "rate limit exceeded" }
-func (e retryAfterErr) StatusCode() int           { return 429 }
-func (e retryAfterErr) RetryAfter() time.Duration { return e.delay }
-
 func TestClassifyErrorDistinguishesOperationalOutcomes(t *testing.T) {
 	tests := []struct {
 		name string
@@ -189,35 +181,3 @@ func TestGenerateWithRetryExposesRetryErrorKind(t *testing.T) {
 		t.Fatalf("retry error does not unwrap provider status: %v", err)
 	}
 }
-
-func TestGenerateWithRetryHonorsRetryAfterWhenLongerThanBackoff(t *testing.T) {
-	attempts := 0
-	model := retryModel{generate: func(context.Context, *Request, ...GenerateOption) (*Response, error) {
-		attempts++
-		if attempts == 1 {
-			return nil, retryAfterErr{delay: 25 * time.Millisecond}
-		}
-		return &Response{Reply: "ok"}, nil
-	}}
-
-	start := time.Now()
-	resp, err := GenerateWithRetry(context.Background(), model, &Request{Prompt: "hi"}, GeneratePolicy{
-		MaxAttempts: 2,
-		Backoff:     time.Millisecond,
-	})
-	if err != nil {
-		t.Fatalf("GenerateWithRetry returned error: %v", err)
-	}
-	if resp.Reply != "ok" {
-		t.Fatalf("reply = %q, want ok", resp.Reply)
-	}
-	if elapsed := time.Since(start); elapsed < 20*time.Millisecond {
-		t.Fatalf("retry delay = %s, want RetryAfter delay to dominate base backoff", elapsed)
-	}
-}
-
-func TestGenerateWithRetryCapsRetryAfter(t *testing.T) {
-	if got := retryBackoff(retryAfterErr{delay: time.Minute}, 1, time.Millisecond); got != 30*time.Second {
-		t.Fatalf("retryBackoff() = %s, want 30s cap", got)
-	}
-}

ai/together/together.go

@@ -22,14 +22,12 @@ import (
 	"strings"
 
 	"go-micro.dev/v6/ai"
-	"go-micro.dev/v6/ai/internal/openaiapi"
 )
 
 func init() {
 	ai.Register("together", func(opts ...ai.Option) ai.Model {
 		return NewProvider(opts...)
 	})
-	ai.RegisterStream("together")
 }
 
 type Provider struct {
@@ -121,7 +119,7 @@ func (p *Provider) Generate(ctx context.Context, req *ai.Request, opts ...ai.Gen
 }
 
 func (p *Provider) Stream(ctx context.Context, req *ai.Request, opts ...ai.GenerateOption) (ai.Stream, error) {
-	return openaiapi.Stream(ctx, p.opts, req, "/v1/chat/completions")
+	return nil, fmt.Errorf("%w: together provider", ai.ErrStreamingUnsupported)
 }
 
 func (p *Provider) callAPI(ctx context.Context, req map[string]any) (*ai.Response, map[string]any, error) {

ai/together/together_test.go

@@ -2,11 +2,7 @@ package together
 
 import (
 	"context"
-	"encoding/json"
 	"errors"
-	"io"
-	"net/http"
-	"net/http/httptest"
 	"testing"
 
 	"go-micro.dev/v6/ai"
@@ -44,44 +40,9 @@ func TestProvider_Generate_NoAPIKey(t *testing.T) {
 	}
 }
 
-func TestProvider_Stream(t *testing.T) {
-	var sawStream bool
-	ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		if r.URL.Path != "/v1/chat/completions" {
-			t.Fatalf("path = %s, want /v1/chat/completions", r.URL.Path)
-		}
-		var body map[string]any
-		if err := json.NewDecoder(r.Body).Decode(&body); err != nil {
-			t.Fatalf("decode request: %v", err)
-		}
-		sawStream, _ = body["stream"].(bool)
-		w.Header().Set("Content-Type", "text/event-stream")
-		_, _ = w.Write([]byte("data: {\"choices\":[{\"delta\":{\"content\":\"hel\"}}]}\n\n"))
-		_, _ = w.Write([]byte("data: {\"choices\":[{\"delta\":{\"content\":\"lo\"}}]}\n\n"))
-		_, _ = w.Write([]byte("data: [DONE]\n\n"))
-	}))
-	defer ts.Close()
-
-	p := NewProvider(ai.WithAPIKey("test-key"), ai.WithBaseURL(ts.URL))
-	stream, err := p.Stream(context.Background(), &ai.Request{Prompt: "Hello"})
-	if err != nil {
-		t.Fatalf("Stream returned error: %v", err)
-	}
-	defer stream.Close()
-	if !sawStream {
-		t.Fatal("stream request did not set stream=true")
-	}
-
-	first, err := stream.Recv()
-	if err != nil || first.Reply != "hel" {
-		t.Fatalf("first chunk = %#v, %v; want hel", first, err)
-	}
-	second, err := stream.Recv()
-	if err != nil || second.Reply != "lo" {
-		t.Fatalf("second chunk = %#v, %v; want lo", second, err)
-	}
-	if _, err := stream.Recv(); !errors.Is(err, io.EOF) {
-		t.Fatalf("final error = %v, want EOF", err)
+func TestProvider_Stream_NotImplemented(t *testing.T) {
+	if _, err := NewProvider().Stream(context.Background(), &ai.Request{Prompt: "hi"}); !errors.Is(err, ai.ErrStreamingUnsupported) {
+		t.Fatalf("Stream error = %v, want ErrStreamingUnsupported", err)
 	}
 }
 

cmd/micro/cli/deploy/deploy.go

@@ -42,28 +42,14 @@ func Deploy(c *cli.Context) error {
 		return showDeployHelp()
 	}
 
-	target, remotePath := resolveDeployTarget(c, target, cfg)
-
-	return deploySSH(c, target, cfg, remotePath)
-}
-
-func resolveDeployTarget(c *cli.Context, target string, cfg *config.Config) (string, string) {
-	remotePath := c.String("path")
-	if remotePath == "" {
-		remotePath = defaultRemotePath
-	}
-
 	// Check if target is a named target from config
 	if cfg != nil {
 		if dt, ok := cfg.Deploy[target]; ok {
 			target = dt.SSH
-			if dt.Path != "" && !c.IsSet("path") {
-				remotePath = dt.Path
-			}
 		}
 	}
 
-	return target, remotePath
+	return deploySSH(c, target, cfg)
 }
 
 func showDeployHelp() error {
@@ -96,7 +82,7 @@ func showDeployTargets(cfg *config.Config) error {
 	return fmt.Errorf("%s", sb.String())
 }
 
-func deploySSH(c *cli.Context, target string, cfg *config.Config, remotePath string) error {
+func deploySSH(c *cli.Context, target string, cfg *config.Config) error {
 	dir := c.Args().Get(1)
 	if dir == "" {
 		dir = "."
@@ -112,6 +98,7 @@ func deploySSH(c *cli.Context, target string, cfg *config.Config, remotePath str
 		cfg, _ = config.Load(absDir)
 	}
 
+	remotePath := c.String("path")
 	if remotePath == "" {
 		remotePath = defaultRemotePath
 	}

cmd/micro/cli/deploy/deploy_test.go

@@ -1,120 +0,0 @@
-package deploy
-
-import (
-	"flag"
-	"os"
-	"strings"
-	"testing"
-
-	"github.com/urfave/cli/v2"
-	"go-micro.dev/v6/cmd/micro/run/config"
-)
-
-func newDeployTestContext(t *testing.T, args ...string) *cli.Context {
-	t.Helper()
-	set := flag.NewFlagSet("deploy", flag.ContinueOnError)
-	set.String("path", defaultRemotePath, "")
-	set.String("ssh", "", "")
-	set.String("service", "", "")
-	set.Bool("build", false, "")
-	if err := set.Parse(args); err != nil {
-		t.Fatalf("parse flags: %v", err)
-	}
-	return cli.NewContext(cli.NewApp(), set, nil)
-}
-
-func TestDeployNoTargetExplainsInitAndDeployHandoff(t *testing.T) {
-	err := showDeployHelp()
-	if err == nil {
-		t.Fatal("expected missing target guidance")
-	}
-	msg := err.Error()
-	for _, want := range []string{
-		"no deployment target specified",
-		"sudo micro init --server",
-		"micro deploy user@your-server",
-		"deploy prod",
-	} {
-		if !strings.Contains(msg, want) {
-			t.Fatalf("missing %q in guidance:\n%s", want, msg)
-		}
-	}
-}
-
-func TestDeployListsConfiguredTargetsWhenNoTargetProvided(t *testing.T) {
-	err := showDeployTargets(&config.Config{Deploy: map[string]*config.DeployTarget{
-		"prod":    {Name: "prod", SSH: "deploy@prod.example.com"},
-		"staging": {Name: "staging", SSH: "deploy@staging.example.com"},
-	}})
-	if err == nil {
-		t.Fatal("expected configured target guidance")
-	}
-	msg := err.Error()
-	for _, want := range []string{
-		"Available deploy targets:",
-		"prod -> deploy@prod.example.com",
-		"staging -> deploy@staging.example.com",
-		"micro deploy <target>",
-	} {
-		if !strings.Contains(msg, want) {
-			t.Fatalf("missing %q in configured target guidance:\n%s", want, msg)
-		}
-	}
-}
-
-func TestResolveDeployTargetUsesConfigTargetAndPath(t *testing.T) {
-	ctx := newDeployTestContext(t, "prod")
-	cfg := &config.Config{Deploy: map[string]*config.DeployTarget{
-		"prod": {Name: "prod", SSH: "deploy@prod.example.com", Path: "/srv/micro"},
-	}}
-
-	target, remotePath := resolveDeployTarget(ctx, ctx.Args().First(), cfg)
-	if target != "deploy@prod.example.com" {
-		t.Fatalf("target = %q, want configured SSH", target)
-	}
-	if remotePath != "/srv/micro" {
-		t.Fatalf("remotePath = %q, want configured path", remotePath)
-	}
-}
-
-func TestResolveDeployTargetAllowsCLIPathOverride(t *testing.T) {
-	ctx := newDeployTestContext(t, "--path", "/tmp/micro", "prod")
-	cfg := &config.Config{Deploy: map[string]*config.DeployTarget{
-		"prod": {Name: "prod", SSH: "deploy@prod.example.com", Path: "/srv/micro"},
-	}}
-
-	target, remotePath := resolveDeployTarget(ctx, ctx.Args().First(), cfg)
-	if target != "deploy@prod.example.com" {
-		t.Fatalf("target = %q, want configured SSH", target)
-	}
-	if remotePath != "/tmp/micro" {
-		t.Fatalf("remotePath = %q, want CLI override", remotePath)
-	}
-}
-
-func TestDeployConfigParserSupportsDeployTargets(t *testing.T) {
-	dir := t.TempDir()
-	path := dir + "/micro.mu"
-	content := `service api
-    path ./api
-
-deploy prod
-    ssh deploy@prod.example.com
-    path /srv/micro
-`
-	if err := os.WriteFile(path, []byte(content), 0644); err != nil {
-		t.Fatalf("write config: %v", err)
-	}
-
-	cfg, err := config.ParseMu(path)
-	if err != nil {
-		t.Fatalf("parse config: %v", err)
-	}
-	prod := cfg.Deploy["prod"]
-	if prod == nil {
-		t.Fatal("missing prod deploy target")
-	}
-	if prod.SSH != "deploy@prod.example.com" || prod.Path != "/srv/micro" {
-		t.Fatalf("deploy target = %#v", prod)
-	}
-}

cmd/micro/inspect/inspect.go

@@ -1,169 +0,0 @@
-// Package inspect registers the 'micro inspect' CLI command.
-package inspect
-
-import (
-	"context"
-	"encoding/json"
-	"fmt"
-	"io"
-	"os"
-
-	"github.com/urfave/cli/v2"
-	goagent "go-micro.dev/v6/agent"
-	"go-micro.dev/v6/cmd"
-	aiflow "go-micro.dev/v6/flow"
-	"go-micro.dev/v6/store"
-)
-
-func init() {
-	cmd.Register(&cli.Command{
-		Name:  "inspect",
-		Usage: "Inspect recent agent and workflow activity",
-		Description: `Inspect is the CLI checkpoint in the local scaffold → run → chat → inspect loop.
-It reads durable local run history, so it works after the agent or flow has stopped.`,
-		Subcommands: []*cli.Command{
-			{
-				Name:      "agent",
-				Usage:     "Show recent recorded runs for an agent",
-				ArgsUsage: "[agent]",
-				Flags:     inspectAgentFlags(),
-				Action:    inspectAgent,
-			},
-			{
-				Name:      "flow",
-				Usage:     "Show durable run history for a flow",
-				ArgsUsage: "[flow]",
-				Flags:     inspectFlowFlags(),
-				Action:    inspectFlow,
-			},
-		},
-	})
-}
-
-func inspectAgentFlags() []cli.Flag {
-	return []cli.Flag{
-		&cli.BoolFlag{Name: "json", Usage: "Print run summaries as JSON for automation"},
-		&cli.StringFlag{Name: "status", Usage: "Only show runs with this status (running, done, error, refused)"},
-		&cli.StringFlag{Name: "trace", Usage: "Only show runs whose trace id matches this full id or prefix"},
-		&cli.IntFlag{Name: "limit", Usage: "Show the most recently updated N runs"},
-	}
-}
-
-func inspectFlowFlags() []cli.Flag {
-	return []cli.Flag{
-		&cli.BoolFlag{Name: "json", Usage: "Print durable run history as JSON for automation"},
-		&cli.BoolFlag{Name: "pending", Usage: "Only show runs that have not completed"},
-		&cli.StringFlag{Name: "status", Usage: "Only show runs with this status (running, done, failed)"},
-		&cli.IntFlag{Name: "limit", Usage: "Show the most recently updated N runs"},
-		&cli.StringFlag{Name: "stage", Usage: "Only show runs currently checkpointed at this stage"},
-	}
-}
-
-func inspectAgent(c *cli.Context) error {
-	name := c.Args().First()
-	if name == "" {
-		return fmt.Errorf("agent name required: micro inspect agent <name>")
-	}
-	opts := goagent.RunListOptions{Status: c.String("status"), TraceID: c.String("trace"), Limit: c.Int("limit")}
-	runs, err := goagent.ListRunSummariesWithOptions(store.DefaultStore, name, opts)
-	if err != nil {
-		return err
-	}
-	return writeAgentInspection(os.Stdout, name, runs, c.Bool("json"))
-}
-
-func writeAgentInspection(w io.Writer, name string, runs []goagent.RunSummary, asJSON bool) error {
-	if asJSON {
-		enc := json.NewEncoder(w)
-		enc.SetIndent("", "  ")
-		return enc.Encode(runs)
-	}
-	if len(runs) == 0 {
-		fmt.Fprintf(w, "  No agent runs recorded for %q. After chatting, try: micro inspect agent %s\n", name, name)
-		return nil
-	}
-	fmt.Fprintf(w, "  Agent %q runs\n", name)
-	for _, run := range runs {
-		fmt.Fprintf(w, "  %s  status=%s  events=%d  last=%s", run.RunID, run.Status, run.Events, run.LastKind)
-		if run.LastError != "" {
-			fmt.Fprintf(w, "  error=%q", run.LastError)
-		}
-		if run.TraceID != "" {
-			fmt.Fprintf(w, "  trace=%s", shortID(run.TraceID))
-		}
-		fmt.Fprintln(w)
-	}
-	return nil
-}
-
-func inspectFlow(c *cli.Context) error {
-	name := c.Args().First()
-	if name == "" {
-		return fmt.Errorf("flow name required: micro inspect flow <name>")
-	}
-	runs, err := aiflow.StoreCheckpoint(nil, name).List(context.Background())
-	if err != nil {
-		return err
-	}
-	runs = filterFlowInspection(runs, c.Bool("pending"), c.String("status"), c.String("stage"), c.Int("limit"))
-	return writeFlowInspection(os.Stdout, name, runs, c.Bool("json"), c.Bool("pending"))
-}
-
-func filterFlowInspection(runs []aiflow.Run, pending bool, status, stage string, limit int) []aiflow.Run {
-	filtered := make([]aiflow.Run, 0, len(runs))
-	for _, run := range runs {
-		if pending && run.Status == "done" {
-			continue
-		}
-		if status != "" && run.Status != status {
-			continue
-		}
-		if stage != "" && run.State.Stage != stage {
-			continue
-		}
-		filtered = append(filtered, run)
-	}
-	if limit > 0 && len(filtered) > limit {
-		return filtered[len(filtered)-limit:]
-	}
-	return filtered
-}
-
-func writeFlowInspection(w io.Writer, name string, runs []aiflow.Run, asJSON, pending bool) error {
-	if asJSON {
-		enc := json.NewEncoder(w)
-		enc.SetIndent("", "  ")
-		return enc.Encode(runs)
-	}
-	if len(runs) == 0 {
-		if pending {
-			fmt.Fprintf(w, "  No pending flow runs recorded for %q.\n", name)
-			return nil
-		}
-		fmt.Fprintf(w, "  No flow runs recorded for %q. After executing a durable flow, try: micro inspect flow %s\n", name, name)
-		return nil
-	}
-	fmt.Fprintf(w, "  Flow %q runs\n", name)
-	for _, run := range runs {
-		stage := run.State.Stage
-		if stage == "" {
-			stage = "-"
-		}
-		fmt.Fprintf(w, "  %s  status=%s  stage=%s  steps=%d", shortID(run.ID), run.Status, stage, len(run.Steps))
-		for _, step := range run.Steps {
-			if step.Error != "" {
-				fmt.Fprintf(w, "  error=%q", step.Error)
-				break
-			}
-		}
-		fmt.Fprintln(w)
-	}
-	return nil
-}
-
-func shortID(id string) string {
-	if len(id) <= 12 {
-		return id
-	}
-	return id[:12]
-}

cmd/micro/inspect/inspect_test.go

@@ -1,64 +0,0 @@
-package inspect
-
-import (
-	"bytes"
-	"encoding/json"
-	"strings"
-	"testing"
-
-	goagent "go-micro.dev/v6/agent"
-	aiflow "go-micro.dev/v6/flow"
-)
-
-func TestWriteAgentInspectionIncludesActionableBreadcrumbs(t *testing.T) {
-	runs := []goagent.RunSummary{{RunID: "run-1", Status: "error", Events: 4, LastKind: "tool", LastError: "boom", TraceID: "1234567890abcdef"}}
-	var out bytes.Buffer
-	if err := writeAgentInspection(&out, "support", runs, false); err != nil {
-		t.Fatal(err)
-	}
-	got := out.String()
-	for _, want := range []string{"Agent \"support\" runs", "run-1", "status=error", "events=4", "last=tool", `error="boom"`, "trace=1234567890ab"} {
-		if !strings.Contains(got, want) {
-			t.Fatalf("output missing %q:\n%s", want, got)
-		}
-	}
-}
-
-func TestWriteAgentInspectionEmptyStateNamesInspectCommand(t *testing.T) {
-	var out bytes.Buffer
-	if err := writeAgentInspection(&out, "support", nil, false); err != nil {
-		t.Fatal(err)
-	}
-	if got := out.String(); !strings.Contains(got, "micro inspect agent support") {
-		t.Fatalf("empty state missing next step: %q", got)
-	}
-}
-
-func TestWriteFlowInspectionIncludesFailedStepBreadcrumb(t *testing.T) {
-	runs := []aiflow.Run{{ID: "1234567890abcdef", Status: "failed", State: aiflow.State{Stage: "charge"}, Steps: []aiflow.StepRecord{{Name: "charge", Status: "failed", Error: "card declined"}}}}
-	var out bytes.Buffer
-	if err := writeFlowInspection(&out, "checkout", runs, false, false); err != nil {
-		t.Fatal(err)
-	}
-	got := out.String()
-	for _, want := range []string{"Flow \"checkout\" runs", "1234567890ab", "status=failed", "stage=charge", "steps=1", `error="card declined"`} {
-		if !strings.Contains(got, want) {
-			t.Fatalf("output missing %q:\n%s", want, got)
-		}
-	}
-}
-
-func TestWriteFlowInspectionJSON(t *testing.T) {
-	runs := []aiflow.Run{{ID: "run-1", Flow: "checkout", Status: "done"}}
-	var out bytes.Buffer
-	if err := writeFlowInspection(&out, "checkout", runs, true, false); err != nil {
-		t.Fatal(err)
-	}
-	var got []aiflow.Run
-	if err := json.Unmarshal(out.Bytes(), &got); err != nil {
-		t.Fatalf("invalid JSON: %v\n%s", err, out.String())
-	}
-	if len(got) != 1 || got[0].ID != "run-1" || got[0].Status != "done" {
-		t.Fatalf("decoded runs = %+v", got)
-	}
-}

cmd/micro/main.go

@@ -12,7 +12,6 @@ import (
 	_ "go-micro.dev/v6/cmd/micro/cli/build"
 	_ "go-micro.dev/v6/cmd/micro/cli/deploy"
 	_ "go-micro.dev/v6/cmd/micro/flow"
-	_ "go-micro.dev/v6/cmd/micro/inspect"
 	_ "go-micro.dev/v6/cmd/micro/mcp"
 	_ "go-micro.dev/v6/cmd/micro/resource"
 	_ "go-micro.dev/v6/cmd/micro/run"

cmd/micro/zero_to_hero_cli_test.go

@@ -19,7 +19,7 @@ func TestZeroToHeroCLIBoundaries(t *testing.T) {
 		}
 	}
 
-	for _, want := range []string{"run", "chat", "flow", "inspect"} {
+	for _, want := range []string{"run", "chat", "flow"} {
 		if !commands[want] {
 			t.Fatalf("missing %q command", want)
 		}
@@ -27,7 +27,4 @@ func TestZeroToHeroCLIBoundaries(t *testing.T) {
 	if !subcommands["flow"]["runs"] {
 		t.Fatal("missing inspect boundary: flow runs")
 	}
-	if !subcommands["inspect"]["agent"] || !subcommands["inspect"]["flow"] {
-		t.Fatal("missing inspect boundary: inspect agent/flow")
-	}
 }

examples/agent-human-input/README.md

@@ -1,36 +0,0 @@
-# Agent Human Input Pause/Resume
-
-Agents can pause a durable run when the model needs a human decision before it
-can continue. This keeps the services → agents → workflows lifecycle in one
-runtime: services expose tools, the agent decides it needs operator input, and
-the same checkpointed run resumes once that input arrives.
-
-## Pattern
-
-```go
-cp := flow.StoreCheckpoint(nil, "deploy-agent")
-ag := agent.New(
-    agent.Name("deploy-agent"),
-    agent.WithCheckpoint(cp),
-)
-
-resp, err := ag.Ask(ctx, "Deploy the service")
-if err != nil {
-    // If the model called the built-in request_input tool, the run is saved as
-    // paused/input-required instead of losing state or completing early.
-    pending, _ := agent.Pending(ctx, ag)
-    runID := pending[0].ID
-
-    // Later, after an operator supplies the missing answer, the same run ID
-    // continues with the original prompt, human input, memory, and completed
-    // tool history intact.
-    resp, err = agent.ResumeInput(ctx, ag, runID, "Deploy to us-east-1")
-}
-_ = resp
-```
-
-The model sees a built-in `request_input` tool with a `prompt` argument. When it
-calls that tool, Go Micro persists the run with status `paused` and stage
-`input-required`. Plain `agent.Resume` continues to support completed, failed,
-and approval-paused runs; input-required runs are resumed with
-`agent.ResumeInput` so the human response is explicit.

examples/support/main.go

@@ -212,39 +212,37 @@ func waitFor(reg registry.Registry, names ...string) {
 	}
 }
 
-func runSupport(provider string) error {
+func main() {
+	provider := flag.String("provider", "mock", "LLM provider: mock (default), anthropic, openai, ...")
+	flag.Parse()
+
 	apiKey := ""
-	if provider == "mock" {
+	if *provider == "mock" {
 		ai.Register("mock", newMock)
-	} else if apiKey = providerKey(provider); apiKey == "" {
-		return fmt.Errorf("no API key for provider %q — set MICRO_AI_API_KEY or the provider's key env", provider)
+	} else if apiKey = providerKey(*provider); apiKey == "" {
+		fmt.Printf("no API key for provider %q — set MICRO_AI_API_KEY or the provider's key env\n", *provider)
+		os.Exit(1)
 	}
 
-	fmt.Printf("\n\033[1mSupport desk (provider: %s)\033[0m\n\n", provider)
+	fmt.Printf("\n\033[1mSupport desk (provider: %s)\033[0m\n\n", *provider)
 
 	// Shared in-memory infrastructure so the demo runs in one process.
 	reg := registry.NewMemoryRegistry()
 	br := broker.NewMemoryBroker()
 	if err := br.Connect(); err != nil {
-		return fmt.Errorf("broker connect: %w", err)
+		fmt.Println("broker connect:", err)
+		os.Exit(1)
 	}
 	cl := client.NewClient(client.Registry(reg), client.Selector(selector.NewSelector(selector.Registry(reg))))
 
 	// Services.
 	tickets := new(TicketService)
 	notify := new(NotifyService)
-	var services []service.Service
 	for name, h := range map[string]any{"customers": new(CustomerService), "tickets": tickets, "notify": notify} {
-		svc := service.New(service.Name(name), service.Address("127.0.0.1:0"), service.Registry(reg), service.Client(cl), service.HandleSignal(false))
+		svc := service.New(service.Name(name), service.Registry(reg), service.Client(cl))
 		_ = svc.Handle(h)
-		services = append(services, svc)
 		go svc.Run()
 	}
-	defer func() {
-		for _, svc := range services {
-			_ = svc.Server().Stop()
-		}
-	}()
 
 	// The support agent manages the three services. The approval gate is
 	// the human-in-the-loop: it can read and triage freely, but emailing a
@@ -252,11 +250,10 @@ func runSupport(provider string) error {
 	// it for a person or a policy; here we approve and log.
 	support := agent.New(
 		agent.Name("support"),
-		agent.Address("127.0.0.1:0"),
 		agent.Services("customers", "tickets", "notify"),
 		agent.Prompt("You are a support agent. For each ticket, look up the customer, set an "+
 			"appropriate priority, and reply to them. Escalate billing issues."),
-		agent.Provider(provider), agent.APIKey(apiKey),
+		agent.Provider(*provider), agent.APIKey(apiKey),
 		agent.ApproveTool(func(tool string, input map[string]any) (bool, string) {
 			if strings.Contains(tool, "Send") {
 				fmt.Printf("  \033[33m▣ approval gate\033[0m %s(%v) — approved\n", tool, input["to"])
@@ -278,7 +275,8 @@ func runSupport(provider string) error {
 		flow.Prompt("A new support ticket arrived: {{.Data}}. Handle it."),
 	)
 	if err := intake.Register(reg, br, cl); err != nil {
-		return fmt.Errorf("flow register: %w", err)
+		fmt.Println("flow register:", err)
+		os.Exit(1)
 	}
 	defer intake.Stop()
 
@@ -289,7 +287,8 @@ func runSupport(provider string) error {
 	fmt.Println("\033[1m> event:\033[0m events.ticket.created", string(body))
 	fmt.Println()
 	if err := br.Publish("events.ticket.created", &broker.Message{Body: body}); err != nil {
-		return fmt.Errorf("publish: %w", err)
+		fmt.Println("publish:", err)
+		os.Exit(1)
 	}
 
 	// Wait for the agent to act.
@@ -306,18 +305,7 @@ func runSupport(provider string) error {
 	}
 	if notify.sent >= 1 {
 		fmt.Println("\n\033[32m✓ ticket triaged and the customer was replied to — triggered by an event\033[0m")
-		return nil
-	}
-	fmt.Println("\n\033[31m✗ the agent did not complete the triage\033[0m")
-	return fmt.Errorf("support agent did not complete triage")
-}
-
-func main() {
-	provider := flag.String("provider", "mock", "LLM provider: mock (default), anthropic, openai, ...")
-	flag.Parse()
-
-	if err := runSupport(*provider); err != nil {
-		fmt.Println(err)
-		os.Exit(1)
+	} else {
+		fmt.Println("\n\033[31m✗ the agent did not complete the triage\033[0m")
 	}
 }

examples/support/main_test.go

@@ -1,9 +0,0 @@
-package main
-
-import "testing"
-
-func TestRunSupportMockSmoke(t *testing.T) {
-	if err := runSupport("mock"); err != nil {
-		t.Fatalf("support example failed: %v", err)
-	}
-}

gateway/mcp/httpjsonrpc.go

@@ -1,89 +0,0 @@
-package mcp
-
-import (
-	"encoding/json"
-	"net/http"
-)
-
-// NewHandler returns an http.Handler serving the MCP protocol over HTTP as
-// JSON-RPC 2.0 (initialize, ping, tools/list, tools/call), backed by the
-// resolver. Mount it on your own server (e.g. POST /mcp): the gateway provides
-// the protocol; you keep your routes, middleware and any human-facing docs page.
-func NewHandler(r Resolver) http.Handler {
-	return http.HandlerFunc(func(w http.ResponseWriter, req *http.Request) {
-		if req.Method != http.MethodPost {
-			http.Error(w, "method not allowed", http.StatusMethodNotAllowed)
-			return
-		}
-		var rpc struct {
-			JSONRPC string          `json:"jsonrpc"`
-			ID      json.RawMessage `json:"id"`
-			Method  string          `json:"method"`
-			Params  json.RawMessage `json:"params"`
-		}
-		if err := json.NewDecoder(req.Body).Decode(&rpc); err != nil {
-			writeRPCError(w, nil, ParseError, "Parse error", err.Error())
-			return
-		}
-		ctx := req.Context()
-		switch rpc.Method {
-		case "initialize":
-			writeRPCResult(w, rpc.ID, map[string]interface{}{
-				"protocolVersion": "2024-11-05",
-				"capabilities":    map[string]interface{}{"tools": map[string]interface{}{}},
-				"serverInfo":      map[string]interface{}{"name": "go-micro-mcp", "version": "1.0.0"},
-			})
-		case "ping":
-			writeRPCResult(w, rpc.ID, map[string]interface{}{})
-		case "tools/list":
-			tools, err := r.List(ctx)
-			if err != nil {
-				writeRPCError(w, rpc.ID, InternalError, "Failed to list tools", err.Error())
-				return
-			}
-			list := make([]map[string]interface{}, 0, len(tools))
-			for _, t := range tools {
-				list = append(list, map[string]interface{}{
-					"name": t.Name, "description": t.Description, "inputSchema": t.InputSchema,
-				})
-			}
-			writeRPCResult(w, rpc.ID, map[string]interface{}{"tools": list})
-		case "tools/call":
-			var p struct {
-				Name      string                 `json:"name"`
-				Arguments map[string]interface{} `json:"arguments"`
-			}
-			if err := json.Unmarshal(rpc.Params, &p); err != nil {
-				writeRPCError(w, rpc.ID, InvalidParams, "Invalid params", err.Error())
-				return
-			}
-			result, err := r.Call(ctx, p.Name, p.Arguments)
-			if err != nil {
-				writeRPCError(w, rpc.ID, InternalError, "Tool call failed", err.Error())
-				return
-			}
-			writeRPCResult(w, rpc.ID, map[string]interface{}{
-				"content": []map[string]interface{}{{"type": "text", "text": result}},
-			})
-		default:
-			writeRPCError(w, rpc.ID, MethodNotFound, "Method not found", rpc.Method)
-		}
-	})
-}
-
-func writeRPCResult(w http.ResponseWriter, id json.RawMessage, result interface{}) {
-	w.Header().Set("Content-Type", "application/json")
-	_ = json.NewEncoder(w).Encode(map[string]interface{}{"jsonrpc": "2.0", "id": rawOrNull(id), "result": result})
-}
-
-func writeRPCError(w http.ResponseWriter, id json.RawMessage, code int, msg, data string) {
-	w.Header().Set("Content-Type", "application/json")
-	_ = json.NewEncoder(w).Encode(map[string]interface{}{"jsonrpc": "2.0", "id": rawOrNull(id), "error": map[string]interface{}{"code": code, "message": msg, "data": data}})
-}
-
-func rawOrNull(id json.RawMessage) interface{} {
-	if len(id) == 0 {
-		return nil
-	}
-	return id
-}

gateway/mcp/resolver.go

@@ -1,118 +0,0 @@
-package mcp
-
-import (
-	"context"
-	"fmt"
-	"sync"
-
-	"go-micro.dev/v6/ai"
-	"go-micro.dev/v6/client"
-	"go-micro.dev/v6/registry"
-)
-
-// ToolFunc executes a manually-registered tool.
-type ToolFunc func(ctx context.Context, args map[string]any) (string, error)
-
-// Resolver supplies the gateway's tools and executes calls. Swapping the
-// resolver changes where tools come from without touching the MCP protocol or
-// transport:
-//
-//   - NewManualResolver:   tools you register explicitly (full product control,
-//     including tools that are not go-micro services, executed via your own
-//     logic — auth, metering, …).
-//   - NewRegistryResolver: tools auto-discovered from registered services.
-//
-// The built-in store/broker tools are intentionally NOT exposed by any
-// resolver — they remain a development convenience on the legacy Serve() path,
-// never a product default.
-type Resolver interface {
-	// List returns the current tool catalog.
-	List(ctx context.Context) ([]Tool, error)
-	// Call executes a tool by name with JSON arguments, returning its text result.
-	Call(ctx context.Context, name string, args map[string]any) (string, error)
-}
-
-// ManualResolver exposes an explicitly-registered set of tools.
-type ManualResolver struct {
-	mu    sync.RWMutex
-	order []Tool
-	funcs map[string]ToolFunc
-}
-
-// NewManualResolver returns an empty manual resolver.
-func NewManualResolver() *ManualResolver {
-	return &ManualResolver{funcs: map[string]ToolFunc{}}
-}
-
-// Add registers (or replaces) a tool and its handler. Returns the resolver for
-// chaining.
-func (m *ManualResolver) Add(t Tool, fn ToolFunc) *ManualResolver {
-	m.mu.Lock()
-	defer m.mu.Unlock()
-	if _, ok := m.funcs[t.Name]; ok {
-		for i := range m.order {
-			if m.order[i].Name == t.Name {
-				m.order[i] = t
-			}
-		}
-	} else {
-		m.order = append(m.order, t)
-	}
-	m.funcs[t.Name] = fn
-	return m
-}
-
-// List returns the registered tools.
-func (m *ManualResolver) List(_ context.Context) ([]Tool, error) {
-	m.mu.RLock()
-	defer m.mu.RUnlock()
-	out := make([]Tool, len(m.order))
-	copy(out, m.order)
-	return out, nil
-}
-
-// Call runs the handler registered for name.
-func (m *ManualResolver) Call(ctx context.Context, name string, args map[string]any) (string, error) {
-	m.mu.RLock()
-	fn, ok := m.funcs[name]
-	m.mu.RUnlock()
-	if !ok {
-		return "", fmt.Errorf("tool not found: %s", name)
-	}
-	return fn(ctx, args)
-}
-
-// RegistryResolver auto-discovers tools from registered go-micro services and
-// executes them over RPC. It exposes only services — never the internal
-// store/broker tools.
-type RegistryResolver struct {
-	tools *ai.Tools
-}
-
-// NewRegistryResolver discovers services from reg and calls them with cl.
-func NewRegistryResolver(reg registry.Registry, cl client.Client) *RegistryResolver {
-	return &RegistryResolver{tools: ai.NewTools(reg, ai.ToolClient(cl))}
-}
-
-// List discovers the current service tools.
-func (r *RegistryResolver) List(_ context.Context) ([]Tool, error) {
-	discovered, err := r.tools.Discover()
-	if err != nil {
-		return nil, err
-	}
-	out := make([]Tool, 0, len(discovered))
-	for _, t := range discovered {
-		out = append(out, Tool{
-			Name:        t.Name,
-			Description: t.Description,
-			InputSchema: map[string]interface{}{"type": "object", "properties": t.Properties},
-		})
-	}
-	return out, nil
-}
-
-// Call executes a discovered service tool.
-func (r *RegistryResolver) Call(ctx context.Context, name string, args map[string]any) (string, error) {
-	res := r.tools.Handler()(ctx, ai.ToolCall{ID: "1", Name: name, Input: args})
-	return res.Content, nil
-}

gateway/mcp/resolver_test.go

@@ -1,58 +0,0 @@
-package mcp
-
-import (
-	"context"
-	"encoding/json"
-	"net/http"
-	"net/http/httptest"
-	"strings"
-	"testing"
-)
-
-func TestManualResolverHandler(t *testing.T) {
-	res := NewManualResolver().Add(
-		Tool{Name: "echo", Description: "echoes text", InputSchema: map[string]interface{}{
-			"type": "object", "properties": map[string]interface{}{"text": map[string]interface{}{"type": "string"}},
-		}},
-		func(_ context.Context, args map[string]interface{}) (string, error) {
-			s, _ := args["text"].(string)
-			return "you said: " + s, nil
-		},
-	)
-	ts := httptest.NewServer(NewHandler(res))
-	defer ts.Close()
-
-	rpc := func(body string) map[string]interface{} {
-		resp, err := http.Post(ts.URL, "application/json", strings.NewReader(body))
-		if err != nil {
-			t.Fatalf("post: %v", err)
-		}
-		defer resp.Body.Close()
-		var out map[string]interface{}
-		json.NewDecoder(resp.Body).Decode(&out)
-		return out
-	}
-
-	// initialize
-	if out := rpc(`{"jsonrpc":"2.0","id":1,"method":"initialize"}`); out["result"] == nil {
-		t.Fatalf("initialize: %v", out)
-	}
-	// tools/list
-	out := rpc(`{"jsonrpc":"2.0","id":2,"method":"tools/list"}`)
-	tools := out["result"].(map[string]interface{})["tools"].([]interface{})
-	if len(tools) != 1 || tools[0].(map[string]interface{})["name"] != "echo" {
-		t.Fatalf("tools/list: %v", out)
-	}
-	// tools/call
-	out = rpc(`{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"echo","arguments":{"text":"hi"}}}`)
-	content := out["result"].(map[string]interface{})["content"].([]interface{})
-	got := content[0].(map[string]interface{})["text"].(string)
-	if got != "you said: hi" {
-		t.Fatalf("tools/call text = %q", got)
-	}
-	// unknown tool -> error
-	out = rpc(`{"jsonrpc":"2.0","id":4,"method":"tools/call","params":{"name":"nope","arguments":{}}}`)
-	if out["error"] == nil {
-		t.Fatalf("expected error for unknown tool, got %v", out)
-	}
-}

internal/docs/PRIORITIES.md

@@ -19,11 +19,9 @@ redirect the loop; direction always wins.
 items the loop can auto-merge): brand/positioning copy, breaking public-API
 changes, architectural rewrites. Those go to the human.
 
-## Developer experience (ranked)
+## Later (ranked)
 
-1. **Emit agent RunInfo as OpenTelemetry spans** ([#3316](https://github.com/micro/go-micro/issues/3316)) — provider-backed streaming just broadened across the OpenAI-compatible path, so the highest-value remaining Next-roadmap seam is making agent execution operable: model calls, tool calls, failures, and run IDs should correlate with the same traces as services and flows.
-2. **Add memory compaction and retrieval for long-running agents** ([#3321](https://github.com/micro/go-micro/issues/3321)) — after run timelines become observable, tackle the leading Later-roadmap gap: bounded, store-backed memory that keeps long agent runs useful without turning Go Micro into a prompt-layer framework.
-3. **Add human-in-the-loop pause and resume for agent workflows** ([#3329](https://github.com/micro/go-micro/issues/3329)) — once agents are observable and memory-bounded, close the next operability gap for scheduled/looping work: durable pending-input states that let humans approve or provide context and then resume the same service/agent/workflow runtime.
+1. **Add agent memory summarization and retrieval** ([#3273](https://github.com/micro/go-micro/issues/3273)) — with A2A resubscribe/input-required handoffs now shipped, the highest-value remaining roadmap gap is durable memory beyond a fixed context buffer: compact older turns/tool results and retrieve relevant prior facts using the existing agent/store primitives so longer-running agents remain coherent without becoming a separate product layer.
 
 _Seeded by Claude Code from the roadmap + open issues; thereafter maintained by the
 architecture-review pass._

internal/harness/provider-conformance/README.md

@@ -1,84 +0,0 @@
-# Provider conformance
-
-This harness keeps the services → agents → workflows lifecycle honest across the
-supported AI providers. It runs the same end-to-end scenarios against each
-configured provider and treats missing provider keys as an explicit skip, so the
-suite is safe for local development, forks, and scheduled CI.
-
-## What it exercises
-
-`go run ./internal/harness/provider-conformance` fans out over the harnesses in
-`internal/harness`:
-
-- `universe` — service discovery plus agent tool calls over the real runtime.
-- `agent-flow` — a workflow event that drives an agent to call services.
-- `plan-delegate` — plan persistence plus agent-to-agent delegation and service
-  calls.
-
-The command also emits the registered provider capability matrix so the run shows
-which providers advertise model, image, video, and streaming support.
-
-## Local usage
-
-Run the deterministic path with no secrets:
-
-```sh
-go run ./internal/harness/provider-conformance -providers mock
-```
-
-Run every live provider that has a key in the environment:
-
-```sh
-go run ./internal/harness/provider-conformance \
-  -summary-json provider-conformance-summary.json \
-  -summary-markdown provider-conformance-summary.md \
-  -capabilities-markdown provider-capabilities.md
-```
-
-Provider keys are read from `MICRO_AI_API_KEY` or the provider-specific variable:
-
-| Provider | Secret / environment variable |
-| --- | --- |
-| Anthropic | `ANTHROPIC_API_KEY` |
-| OpenAI | `OPENAI_API_KEY` |
-| Gemini | `GEMINI_API_KEY` |
-| Groq | `GROQ_API_KEY` |
-| Mistral | `MISTRAL_API_KEY` |
-| Together | `TOGETHER_API_KEY` |
-| AtlasCloud | `ATLASCLOUD_API_KEY` |
-
-Use `-require-configured` when you want a selected provider without a key to fail
-instead of skip:
-
-```sh
-go run ./internal/harness/provider-conformance \
-  -providers anthropic,openai \
-  -require-configured
-```
-
-## Scheduled CI behavior
-
-The `Harness (E2E)` workflow runs on pushes and pull requests with deterministic
-mock LLMs. On the daily schedule and manual dispatch it also runs the live
-provider conformance job. That job:
-
-1. reads the provider keys from repository secrets,
-2. skips providers whose secrets are absent,
-3. fails when any configured provider fails a harness, and
-4. uploads JSON and Markdown coverage artifacts for the run.
-
-The job also appends the Markdown summary and capability matrix to the GitHub
-Actions step summary, making configured, skipped, and failed provider coverage
-visible without downloading artifacts.
-
-## Adding a provider
-
-To bring a new provider into scheduled conformance:
-
-1. register its `ai` provider implementation and capability metadata,
-2. add the provider name and key variable to `providerEnv` in `main.go`,
-3. import the provider package in `main.go`,
-4. pass the matching repository secret through `.github/workflows/harness.yml`,
-   and
-5. run `go run ./internal/harness/provider-conformance -providers <name> \
-   -require-configured` with a live key before opening the change.

internal/harness/zero-to-hero-ci/README.md

@@ -8,9 +8,8 @@ scripted so CI can run it on every push without external services or model keys.
 
 1. **Run** — `micro run` remains available as the local development entry point.
 2. **Chat** — `micro chat` remains available as the interactive agent entry point.
-3. **Inspect** — `micro inspect agent <name>` and `micro inspect flow <name>`
-   remain available as the local run-history inspection step, with `micro flow
-   runs` preserving durable workflow history inspection.
+3. **Inspect** — `micro flow runs` remains available for durable workflow run
+   history inspection.
 
 After the CLI boundary smoke checks, the script runs the deterministic harnesses
 that boot real services, agents, workflows, store-backed run history, and A2A

internal/website/docs/guides/agent-harness.md

@@ -36,46 +36,13 @@ your stack — the harness *is* the stack.
 | Discovery & RPC | Registry + client; agents and services find and call each other | Shipped |
 | Interop | MCP (tools), A2A (agents), x402 (paid tools) | Shipped |
 | Resilience | Per-call timeout with context propagation; opt-in retry/backoff (`ModelRetry`) across the loop | Shipped |
-| Durable runs | Checkpoint and resume an agent run with the same checkpoint backend flows use | Shipped |
+| Durable runs | Checkpoint and resume an agent run (flows already do) | In progress |
 | Observability | `RunInfo` → OpenTelemetry spans for runs, model calls, tools, delegation, and failures; persisted run history | Shipped |
 | Streaming | `ai.Stream` through chat, agent, and A2A | In progress |
 
 The "in progress" rows are exactly the roadmap's [Now and Next](/docs/roadmap.html),
 and the work is happening in the open.
 
-## Durable agent runs
-
-Agents can persist their execution history to the same `Checkpoint` backend as
-flows. A checkpointed `Ask` records the run id, original prompt, model result,
-and completed tool calls. If the process restarts after a tool succeeds but
-before the model finishes, `AgentResume` continues the same run and returns the
-recorded tool result instead of re-running the side effect. If a run already
-completed, resume returns the persisted response without calling the model.
-
-```go
-agent := micro.NewAgent("conductor",
-    micro.AgentProvider("anthropic"),
-    micro.AgentWithCheckpoint(checkpoint),
-)
-
-resp, err := agent.Ask(ctx, "charge order 42 and send a receipt")
-if err != nil {
-    // On startup, or after a transient failure, discover unfinished work:
-    pending, _ := micro.AgentPending(ctx, agent)
-    for _, run := range pending {
-        _, _ = micro.AgentResume(ctx, agent, run.ID)
-    }
-}
-_ = resp
-```
-
-For human-in-the-loop runs that pause through the built-in `request_input` tool,
-resume with the operator's response:
-
-```go
-_, err := micro.AgentResumeInput(ctx, agent, runID, "Deploy to us-east-1")
-```
-
 ## Observing agent runs
 
 Pass an OpenTelemetry tracer provider when you construct an agent to turn the

internal/website/docs/guides/ai-provider-guide.md

@@ -39,12 +39,12 @@ The built-in providers currently register these capability interfaces:
 | Provider | Chat/text (`ai.Model`) | Image (`ai.ImageModel`) | Video (`ai.VideoModel`) | Streaming (`ai.Stream`) |
 | --- | --- | --- | --- | --- |
 | `anthropic` | Yes | No | No | No |
-| `atlascloud` | Yes | Yes | Yes | Yes |
+| `atlascloud` | Yes | Yes | Yes | No |
 | `gemini` | Yes | No | No | No |
-| `groq` | Yes | No | No | Yes |
-| `mistral` | Yes | No | No | Yes |
+| `groq` | Yes | No | No | No |
+| `mistral` | Yes | No | No | No |
 | `openai` | Yes | Yes | No | Yes |
-| `together` | Yes | No | No | Yes |
+| `together` | Yes | No | No | No |
 
 ## Step 1: Implement the `ai.Model` Interface
 

internal/website/docs/roadmap.md

@@ -36,13 +36,14 @@ The priority is that what exists works everywhere, under real conditions.
 
 ## Next — agentic depth
 
+- **Durable agent loop.** Flows resume; the agent's own loop does not yet. Reuse `Checkpoint` so a long-running agent survives a restart and continues.
 - **Streaming.** Broaden provider-backed `ai.Stream` coverage and keep chat plus A2A `message/stream` working end to end for real chat and long-task UX.
 - **Agent observability.** Wire the new `RunInfo` into OpenTelemetry spans so a run — steps, tool calls, delegation — is traceable. This is also what anyone running it in production will need.
 
 ## Later
 
 - **Memory management** — summarization and retrieval (RAG) beyond a fixed buffer.
-- **Human-in-the-loop** — broaden pause/resume UX around `input-required` runs and approvals.
+- **Human-in-the-loop** — pause and resume mid-run (`input-required`), beyond the binary `ApproveTool` gate.
 - **A2A** — richer live-stream reconnection (`tasks/resubscribe`) and `input-required` handoffs.
 
 ## Developer experience (ongoing)

micro.go

@@ -23,18 +23,12 @@ type Service = service.Service
 // Agent is the interface for an AI agent that manages services.
 type Agent = agent.Agent
 
-// AgentResponse is what an agent returns from Ask or a resumed run.
-type AgentResponse = agent.Response
-
 // AgentOption configures an Agent.
 type AgentOption = agent.Option
 
 // Flow is an event-driven LLM orchestration unit.
 type Flow = flow.Flow
 
-// FlowRun is a checkpointed flow or agent run record.
-type FlowRun = flow.Run
-
 // FlowOption configures a Flow.
 type FlowOption = flow.Option
 
@@ -170,26 +164,6 @@ func AgentWrapTool(w ...ai.ToolWrapper) AgentOption {
 // tool calls, delegation, and failures.
 func AgentTraceProvider(tp trace.TracerProvider) AgentOption { return agent.TraceProvider(tp) }
 
-// AgentWithCheckpoint sets the durability backend for agent Ask runs.
-// It uses the same Checkpoint interface as flows so services, agents,
-// and workflows can share one execution history backend.
-func AgentWithCheckpoint(c Checkpoint) AgentOption { return agent.WithCheckpoint(c) }
-
-// AgentPending returns checkpointed agent runs that have not completed.
-// Use it at process startup to discover agent work that should be resumed.
-func AgentPending(ctx context.Context, a Agent) ([]FlowRun, error) { return agent.Pending(ctx, a) }
-
-// AgentResume resumes a checkpointed agent run by id. Completed runs return
-// the persisted response without calling the model or replaying tool calls.
-func AgentResume(ctx context.Context, a Agent, runID string) (*AgentResponse, error) {
-	return agent.Resume(ctx, a, runID)
-}
-
-// AgentResumeInput resumes a checkpointed agent run waiting for human input.
-func AgentResumeInput(ctx context.Context, a Agent, runID, input string) (*AgentResponse, error) {
-	return agent.ResumeInput(ctx, a, runID, input)
-}
-
 // NewFlow creates an event-driven LLM orchestration unit.
 //
 //	f := micro.NewFlow("onboard-user",