a06f331eb8
CI / benchmark (push) Has been skipped
install-script / posix-syntax (push) Successful in 6m1s
CI / build-onnx (push) Failing after 6m43s
init-smoke / dry-run (push) Failing after 15m57s
security / govulncheck (push) Has been cancelled
security / trivy-fs (push) Has been cancelled
CI / test (1.26, ubuntu-latest) (push) Has been cancelled
Scorecard supply-chain security / Scorecard analysis (push) Has been cancelled
CI / test (1.26, macos-latest) (push) Has been cancelled
CI / build-windows (push) Has been cancelled
CI / lint (push) Has been cancelled
install-script / powershell-syntax (push) Has been cancelled
install-script / install (macos-14) (push) Has been cancelled
install-script / install (ubuntu-latest) (push) Has been cancelled
138 lines
5.1 KiB
Go
138 lines
5.1 KiB
Go
// Package provider is the llm.Provider factory. It is the one place
|
|
// that imports every concrete provider implementation; everything else
|
|
// (the svc layer, the agent loop, the cmd demos) depends only on the
|
|
// llm.Provider interface and calls New here.
|
|
//
|
|
// The build-tag split is contained entirely within the `local`
|
|
// subpackage: provider.New compiles in every build, and selecting
|
|
// "local" without a `-tags llama` binary surfaces as a runtime error
|
|
// from local.New, not a compile failure.
|
|
package provider
|
|
|
|
import (
|
|
"fmt"
|
|
"net/http"
|
|
"os"
|
|
"strings"
|
|
"time"
|
|
|
|
"github.com/zzet/gortex/internal/llm"
|
|
"github.com/zzet/gortex/internal/llm/provider/anthropic"
|
|
"github.com/zzet/gortex/internal/llm/provider/azure"
|
|
"github.com/zzet/gortex/internal/llm/provider/bedrock"
|
|
"github.com/zzet/gortex/internal/llm/provider/claudecli"
|
|
"github.com/zzet/gortex/internal/llm/provider/codex"
|
|
"github.com/zzet/gortex/internal/llm/provider/copilot"
|
|
"github.com/zzet/gortex/internal/llm/provider/cursor"
|
|
"github.com/zzet/gortex/internal/llm/provider/deepseek"
|
|
"github.com/zzet/gortex/internal/llm/provider/gemini"
|
|
"github.com/zzet/gortex/internal/llm/provider/local"
|
|
"github.com/zzet/gortex/internal/llm/provider/ollama"
|
|
"github.com/zzet/gortex/internal/llm/provider/openai"
|
|
"github.com/zzet/gortex/internal/llm/provider/openaicompat"
|
|
"github.com/zzet/gortex/internal/llm/provider/opencode"
|
|
)
|
|
|
|
// New builds the llm.Provider selected by cfg.Provider. cfg should
|
|
// already have defaults applied (see llm.Config.ApplyDefaults) — the
|
|
// HTTP providers rely on the defaulted model / endpoint / key-env
|
|
// values. Returns an error when the provider is unknown or
|
|
// misconfigured (missing model, unset API key) or, for "local", when
|
|
// the binary was built without `-tags llama`; for "claudecli" /
|
|
// "codex", when the `claude` / `codex` binary is not on $PATH.
|
|
func New(cfg llm.Config) (llm.Provider, error) {
|
|
switch cfg.ProviderName() {
|
|
case "local":
|
|
return local.New(cfg.Local)
|
|
case "anthropic":
|
|
ac := cfg.Anthropic
|
|
return anthropic.New(ac.RemoteConfig,
|
|
anthropic.WithPromptCaching(ac.PromptCaching, ac.CacheTTL),
|
|
anthropic.WithThinking(ac.ThinkingMode, ac.ThinkingBudgetTokens, ac.ThinkingDisplay),
|
|
)
|
|
case "openai":
|
|
return openai.New(cfg.OpenAI)
|
|
case "azure":
|
|
return azure.New(cfg.Azure)
|
|
case "ollama":
|
|
return ollama.New(cfg.Ollama)
|
|
case "claudecli":
|
|
return claudecli.New(cfg.ClaudeCLI)
|
|
case "codex":
|
|
return codex.New(cfg.Codex)
|
|
case "copilot":
|
|
return copilot.New(cfg.Copilot)
|
|
case "cursor":
|
|
return cursor.New(cfg.Cursor)
|
|
case "opencode":
|
|
return opencode.New(cfg.Opencode)
|
|
case "gemini":
|
|
return gemini.New(cfg.Gemini)
|
|
case "bedrock":
|
|
return bedrock.New(cfg.Bedrock)
|
|
case "deepseek":
|
|
return deepseek.New(cfg.DeepSeek)
|
|
default:
|
|
if cp, ok := cfg.Custom[cfg.ProviderName()]; ok {
|
|
return newCustom(cfg.ProviderName(), cp)
|
|
}
|
|
return nil, fmt.Errorf("llm: unknown provider %q (want local|anthropic|openai|azure|ollama|claudecli|codex|copilot|cursor|opencode|gemini|bedrock|deepseek, or a registered custom provider)", cfg.ProviderName())
|
|
}
|
|
}
|
|
|
|
// newCustom builds a registered OpenAI-compatible custom provider from
|
|
// its CustomProvider entry. The endpoint is BaseURL + "/chat/completions";
|
|
// the bearer key (when APIKeyEnv is set) and any extra headers are
|
|
// applied to every request; the structured-output strategy comes from
|
|
// SchemaMode. Name() reports "custom" (frontier prompt tier) while the
|
|
// httpx/error tag carries the specific "custom:<name>".
|
|
func newCustom(name string, cp llm.CustomProvider) (llm.Provider, error) {
|
|
base := strings.TrimRight(strings.TrimSpace(cp.BaseURL), "/")
|
|
if base == "" {
|
|
return nil, fmt.Errorf("llm: custom provider %q has no base_url", name)
|
|
}
|
|
if !strings.HasPrefix(base, "http://") && !strings.HasPrefix(base, "https://") {
|
|
return nil, fmt.Errorf("llm: custom provider %q base_url must be http(s), got %q", name, base)
|
|
}
|
|
if strings.TrimSpace(cp.Model) == "" {
|
|
return nil, fmt.Errorf("llm: custom provider %q has no model", name)
|
|
}
|
|
|
|
headers := make(map[string]string, len(cp.Headers)+1)
|
|
for k, v := range cp.Headers {
|
|
headers[k] = v
|
|
}
|
|
if env := strings.TrimSpace(cp.APIKeyEnv); env != "" {
|
|
key := strings.TrimSpace(os.Getenv(env))
|
|
if key == "" {
|
|
return nil, fmt.Errorf("llm: custom provider %q API key env %q is not set", name, env)
|
|
}
|
|
headers["authorization"] = "Bearer " + key
|
|
}
|
|
|
|
var mode openaicompat.SchemaMode
|
|
switch strings.ToLower(strings.TrimSpace(cp.SchemaMode)) {
|
|
case "", "json_schema":
|
|
mode = openaicompat.SchemaJSONSchema
|
|
case "json_object":
|
|
mode = openaicompat.SchemaJSONObject
|
|
case "prompt", "prompt_only", "none":
|
|
mode = openaicompat.SchemaPromptOnly
|
|
default:
|
|
return nil, fmt.Errorf("llm: custom provider %q unknown schema_mode %q (want json_schema|json_object|prompt)", name, cp.SchemaMode)
|
|
}
|
|
|
|
return &openaicompat.Client{
|
|
ProviderID: "custom",
|
|
Tag: "custom:" + name,
|
|
Model: cp.Model,
|
|
URL: base + "/chat/completions",
|
|
Headers: headers,
|
|
HTTPClient: &http.Client{Timeout: 120 * time.Second},
|
|
SchemaMode: mode,
|
|
MaxTokensField: cp.MaxTokensField,
|
|
Temperature: cp.Temperature,
|
|
ReasoningEffort: cp.ReasoningEffort,
|
|
}, nil
|
|
}
|