wehub-resource-sync
112 lines
3.6 KiB
TypeScript
112 lines
3.6 KiB
TypeScript
import 'server-only';
|
|
import { WorkflowNode, WorkflowState } from '../types';
|
|
import { substituteVariables } from '../variable-substitution';
|
|
|
|
/**
|
|
* Execute Extract Node - Uses LLM with JSON schema to extract structured data
|
|
* Server-side only - called from API routes
|
|
*/
|
|
export async function executeExtractNode(
|
|
node: WorkflowNode,
|
|
state: WorkflowState,
|
|
apiKeys?: { anthropic?: string; groq?: string; openai?: string; firecrawl?: string }
|
|
): Promise<any> {
|
|
const { data } = node;
|
|
|
|
try {
|
|
// Substitute variables in instructions
|
|
const instructions = substituteVariables(data.instructions || 'Extract information from the input', state);
|
|
|
|
// Build context from previous node output
|
|
const lastOutput = state.variables?.lastOutput;
|
|
|
|
// Validate API keys are provided
|
|
if (!apiKeys) {
|
|
throw new Error('API keys are required for server-side execution');
|
|
}
|
|
|
|
// Server-side execution only
|
|
const OpenAI = (await import('openai')).default;
|
|
const client = new OpenAI({ apiKey: apiKeys?.openai });
|
|
|
|
// Build full prompt with context
|
|
let fullPrompt = instructions;
|
|
if (lastOutput) {
|
|
const contextData = typeof lastOutput === 'string'
|
|
? lastOutput
|
|
: JSON.stringify(lastOutput, null, 2);
|
|
fullPrompt = `${fullPrompt}\n\nData to extract from:\n${contextData.substring(0, 10000)}`;
|
|
}
|
|
|
|
// Parse JSON schema
|
|
const schema = typeof data.jsonSchema === 'string'
|
|
? JSON.parse(data.jsonSchema)
|
|
: data.jsonSchema;
|
|
|
|
// If MCP tools are configured, use Responses API
|
|
if (data.mcpTools && data.mcpTools.length > 0) {
|
|
const tools = data.mcpTools.map((mcp: any) => ({
|
|
type: 'mcp' as const,
|
|
server_label: mcp.name,
|
|
server_url: mcp.url.includes('{FIRECRAWL_API_KEY}')
|
|
? mcp.url.replace('{FIRECRAWL_API_KEY}', apiKeys?.firecrawl || '')
|
|
: mcp.url,
|
|
authorization: mcp.accessToken ? `Bearer ${mcp.accessToken}` : undefined,
|
|
require_approval: 'never' as const,
|
|
}));
|
|
|
|
const response = await client.responses.create({
|
|
model: 'gpt-4.1',
|
|
tools,
|
|
input: fullPrompt,
|
|
text: {
|
|
format: {
|
|
type: 'json_schema',
|
|
name: 'extraction',
|
|
schema,
|
|
strict: true,
|
|
},
|
|
},
|
|
});
|
|
|
|
const extractedData = JSON.parse(response.output_text || '{}');
|
|
|
|
return {
|
|
extractedData,
|
|
model: 'gpt-4.1',
|
|
tokensUsed: response.usage?.total_tokens || 0,
|
|
mcpToolsUsed: response.output.filter((item: any) => item.type === 'mcp_call').length,
|
|
__variableUpdates: { lastOutput: extractedData }, // Return as separate field for reducer
|
|
};
|
|
}
|
|
|
|
// No MCP - use regular Chat Completions with JSON mode
|
|
const completion = await client.chat.completions.create({
|
|
model: data.model || 'gpt-5-mini',
|
|
messages: [
|
|
{ role: 'user', content: fullPrompt },
|
|
],
|
|
response_format: {
|
|
type: 'json_schema',
|
|
json_schema: {
|
|
name: 'extraction',
|
|
schema,
|
|
strict: true,
|
|
},
|
|
},
|
|
});
|
|
|
|
const extractedData = JSON.parse(completion.choices[0].message.content || '{}');
|
|
|
|
return {
|
|
extractedData,
|
|
model: data.model,
|
|
tokensUsed: completion.usage?.total_tokens || 0,
|
|
__variableUpdates: { lastOutput: extractedData }, // Return as separate field for reducer
|
|
};
|
|
} catch (error) {
|
|
console.error('Extract execution error:', error);
|
|
throw new Error(`Failed to execute extract: ${error instanceof Error ? error.message : 'Unknown error'}`);
|
|
}
|
|
}
|