File size: 4,576 Bytes
b1df986 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 | import OpenAI from 'openai'
import { createLogger } from '../../shared/logging/logger'
import { cleanManimCode } from '../../utils/manim-code-cleaner'
import { getClient } from './client'
import type { CodeRetryContext, RetryCheckpoint } from './types'
import { buildVisionUserMessage } from '../generation/multimodal-input'
import { buildRetryPrompt, getCodeRetrySystemPrompt } from './prompt-builder'
import { dedupeSharedBlocksInMessages } from '../prompt-dedup'
import { createChatCompletionText } from '../openai-stream'
import { buildCompletionParams, readMaxCompletionTokens } from '../../infrastructure/ai-provider/completion-policy'
import { applyPatchSetToCode, extractTargetLine } from './utils'
import { getApiRagService } from '../../knowledge/api-rag'
import { runRepairSession } from './repair-session'
import { inferAutomaticApiRequests } from './api-hints'
const logger = createLogger('CodeRetryCodeGen')
const AI_TEMPERATURE = parseFloat(process.env.AI_TEMPERATURE || '0.7')
const MAX_TOKENS = readMaxCompletionTokens()
function getModel(customApiConfig?: unknown): string {
const model = (customApiConfig as { model?: string } | undefined)?.model
const trimmed = model?.trim() || ''
if (!trimmed) {
throw new Error('No model available')
}
return trimmed
}
export async function retryCodeGeneration(
context: CodeRetryContext,
errorMessage: string,
attempt: number,
currentCode: string,
codeSnippet: string | undefined,
customApiConfig?: unknown,
onCheckpoint?: RetryCheckpoint
): Promise<string> {
const client = getClient(customApiConfig as any)
if (!client) {
throw new Error('No upstream AI is configured for this request')
}
const retryPrompt = buildRetryPrompt(context, errorMessage, attempt, currentCode, codeSnippet)
const apiRagService = getApiRagService()
try {
const automaticRequests = inferAutomaticApiRequests(currentCode, errorMessage)
const automaticResults = await Promise.all(
automaticRequests.map((request) => apiRagService.search('manim', request))
)
const automaticApiContext = automaticResults.length > 0
? `\n\n[[AUTOMATIC_API_CONTEXT]]\n${automaticResults
.map((result) => JSON.stringify(result, null, 2))
.join('\n\n---\n\n')}\n[[END]]\nUse this verified API context. If it is insufficient, return an API_REQUEST.`
: ''
logger.info('Automatic retry API context resolved', {
attempt,
queries: automaticRequests.map((request) => request.query),
candidates: automaticResults.map((result) => result.candidates.length)
})
const requestMessages: OpenAI.Chat.Completions.ChatCompletionMessageParam[] = dedupeSharedBlocksInMessages(
[
{ role: 'system', content: getCodeRetrySystemPrompt(context.promptOverrides) },
{ role: 'user', content: `${retryPrompt}${automaticApiContext}` }
],
context.promptOverrides
).map((message) => message.role === 'user'
? { role: 'user', content: buildVisionUserMessage(message.content, context.referenceImages) }
: message)
const session = await runRepairSession({
messages: requestMessages,
apiRagService,
packageName: 'manim',
concept: context.concept,
attempt,
onCheckpoint,
requestTurn: (messages, turn) => createChatCompletionText(
client,
{
model: getModel(customApiConfig),
messages,
temperature: AI_TEMPERATURE,
...buildCompletionParams(MAX_TOKENS, undefined, client)
},
{ fallbackToNonStream: true, usageLabel: `retry-${attempt}-turn-${turn}` }
)
})
const patchedCode = applyPatchSetToCode(
currentCode,
session.patchSet,
extractTargetLine(errorMessage)
)
const cleaned = cleanManimCode(patchedCode)
logger.info('Code retry patch applied', {
concept: context.concept,
attempt,
turn: session.turn,
mode: session.mode,
patchCount: session.patchSet.patches.length,
codeLength: cleaned.code.length,
patchLengths: session.patchSet.patches.map((patch) => ({
originalSnippetLength: patch.originalSnippet.length,
replacementSnippetLength: patch.replacementSnippet.length
})),
codePreview: cleaned.code.slice(0, 500)
})
return cleaned.code
} catch (error) {
if (error instanceof OpenAI.APIError) {
logger.error('OpenAI API error during code retry', {
attempt,
status: error.status,
message: error.message
})
}
throw error
}
}
|