Skip to content

The Postgres development platform. Supabase gives you a dedicated Postgres database to build your web, mobile, and AI applications.

9.4 KB
Raw Blame
import pgMeta from '@supabase/pg-meta'
import type { JwtPayload } from '@supabase/supabase-js'
import { pipeUIMessageStreamToResponse, safeValidateUIMessages, toUIMessageStream } from 'ai'
import { IS_PLATFORM } from 'common'
import type { NextApiRequest, NextApiResponse } from 'next'
import z from 'zod'

import { executeSql } from '@/data/sql/execute-sql-mutation'
import type { AiOptInLevel } from '@/hooks/misc/useOrgOptedIntoAi'
import { getAIDetails } from '@/lib/ai/ai-details'
import { NO_SCHEMA_ACCESS_MESSAGE } from '@/lib/ai/assistant-context'
import {
  assistantMessageMetadataSchema,
  messagesIncludeLogsSnippets,
  type AssistantMessageMetadata,
} from '@/lib/ai/assistant-message-metadata'
import { ASSISTANT_TIMEOUT_MS } from '@/lib/ai/assistant-timeout'
import { isTracingAllowed } from '@/lib/ai/braintrust-logger'
import { generateAssistantResponse } from '@/lib/ai/generate-assistant-response'
import { isExplorerEnabled } from '@/lib/ai/is-explorer-enabled'
import { getModel } from '@/lib/ai/model'
import {
  DEFAULT_ASSISTANT_BASE_MODEL_ID,
  getAssistantModelEntry,
  isAssistantBaseModelId,
  isKnownAssistantModelId,
  type AssistantModelId,
} from '@/lib/ai/model.utils'
import { getTools } from '@/lib/ai/tools'
import { encodeNotebookToolError } from '@/lib/ai/tools/notebook-tool-error'
import { apiWrapper } from '@/lib/api/apiWrapper'
import { executeQuery } from '@/lib/api/self-hosted/query'
import { getBasePathURL } from '@/lib/helpers'
import { trustedUserEmail } from '@/lib/server/configcat'

export const maxDuration = 300

export const config = {
  api: {
    bodyParser: {
      sizeLimit: '5mb',
    },
  },
}

async function handler(req: NextApiRequest, res: NextApiResponse, claims?: JwtPayload) {
  const { method } = req

  switch (method) {
    case 'POST':
      return handlePost(req, res, claims)
    default:
      res.setHeader('Allow', ['POST'])
      res.status(405).json({
        data: null,
        error: { message: `Method ${method} Not Allowed` },
      })
  }
}

const wrapper = (req: NextApiRequest, res: NextApiResponse) =>
  apiWrapper(req, res, handler, { withAuth: true })

export default wrapper

const requestBodySchema = z.object({
  messages: z.array(z.any()),
  projectRef: z.string(),
  connectionString: z.string(),
  schema: z.string().optional(),
  table: z.string().optional(),
  chatId: z.string().optional(),
  chatName: z.string().optional(),
  supportMode: z.boolean().optional(),
  orgSlug: z.string().optional(),
  model: z.string().optional(),
})

async function handlePost(req: NextApiRequest, res: NextApiResponse, claims?: JwtPayload) {
  const requestStartedAt = Date.now()
  const authorization = req.headers.authorization
  const accessToken = authorization?.replace('Bearer ', '')

  if (IS_PLATFORM && !accessToken) {
    return res.status(401).json({ error: 'Authorization token is required' })
  }

  const userId = claims?.sub

  const body = typeof req.body === 'string' ? JSON.parse(req.body) : req.body
  const { data, error: parseError } = requestBodySchema.safeParse(body)

  if (parseError) {
    return res.status(400).json({ error: 'Invalid request body', issues: parseError.issues })
  }

  const {
    messages: rawMessages,
    projectRef,
    connectionString,
    orgSlug: rawOrgSlug,
    chatId,
    chatName,
    model: rawRequestedModel,
    supportMode,
  } = data

  const requestedModel: AssistantModelId | undefined =
    rawRequestedModel && isKnownAssistantModelId(rawRequestedModel) ? rawRequestedModel : undefined

  const messagesValidation = await safeValidateUIMessages({
    messages: rawMessages,
    metadataSchema: assistantMessageMetadataSchema,
  })
  if (!messagesValidation.success) {
    return res.status(400).json({
      error: 'Invalid request body',
      message: messagesValidation.error.message,
    })
  }
  const messages = messagesValidation.data

  const includesLogsSnippets = messagesIncludeLogsSnippets(messages)

  let aiOptInLevel: AiOptInLevel = 'disabled'
  let hasAccessToAdvanceModel = false
  let projectRegion: string | undefined
  let isHighComplianceProject: boolean | undefined
  let orgId: number | undefined
  let orgSlug: string | undefined
  let planId: string | undefined

  if (!IS_PLATFORM) {
    aiOptInLevel = 'schema'
    hasAccessToAdvanceModel = true
  }

  if (IS_PLATFORM && rawOrgSlug && authorization && projectRef) {
    try {
      const aiDetails = await getAIDetails({ orgSlug: rawOrgSlug, projectRef, authorization })

      aiOptInLevel = aiDetails.aiOptInLevel
      hasAccessToAdvanceModel = aiDetails.hasAccessToAdvanceModel
      orgId = aiDetails.orgId
      orgSlug = aiDetails.orgSlug
      planId = aiDetails.planId
      projectRegion = aiDetails.region
      isHighComplianceProject = aiDetails.isHighComplianceProject
    } catch (error) {
      return res.status(400).json({
        error: 'There was an error fetching your organization details',
      })
    }
  }

  const explorerEnabled = await isExplorerEnabled(trustedUserEmail(claims?.email))

  const envThrottled = process.env.IS_THROTTLED !== 'false'

  let effectiveModel: AssistantModelId = requestedModel ?? DEFAULT_ASSISTANT_BASE_MODEL_ID
  if (!isAssistantBaseModelId(effectiveModel) && (!hasAccessToAdvanceModel || envThrottled)) {
    effectiveModel = DEFAULT_ASSISTANT_BASE_MODEL_ID
  }

  const {
    modelParams,
    error: modelError,
    systemProviderOptions,
  } = await getModel({
    provider: 'openai',
    modelEntry: getAssistantModelEntry(effectiveModel),
  })

  if (modelError) {
    return res.status(500).json({ error: modelError.message })
  }

  try {
    const abortController = new AbortController()
    req.on('close', () => abortController.abort())
    req.on('aborted', () => abortController.abort())
    // Fires when the connection drops. Aborting tears down the remote MCP connection opened
    // in getTools. The TanStack adapter doesn't emit it, so settling the pipe below also aborts.
    res.on('close', () => abortController.abort())

    const tools = await getTools({
      projectRef,
      connectionString,
      authorization,
      aiOptInLevel,
      accessToken,
      baseUrl: getBasePathURL(),
      supportMode,
      isExplorerEnabled: explorerEnabled,
      signal: abortController.signal,
    })

    // Get a list of all schemas to add to context
    const getSchemas = async (): Promise<string> => {
      const pgMetaSchemasList = pgMeta.schemas.list()
      type Schemas = z.infer<(typeof pgMetaSchemasList)['zod']>

      const { result: schemas } = await executeSql<Schemas>(
        {
          projectRef,
          connectionString,
          sql: pgMetaSchemasList.sql,
        },
        undefined,
        {
          'Content-Type': 'application/json',
          ...(authorization && { Authorization: authorization }),
        },
        IS_PLATFORM ? undefined : executeQuery
      )

      return schemas?.length > 0
        ? `The available database schema names are: ${JSON.stringify(schemas)}`
        : NO_SCHEMA_ACCESS_MESSAGE
    }

    const result = await generateAssistantResponse({
      messages,
      ...modelParams,
      tools,
      aiOptInLevel,
      getSchemas: aiOptInLevel !== 'disabled' ? getSchemas : undefined,
      projectRef,
      chatId,
      chatName,
      allowTracing: isTracingAllowed({ projectRegion }),
      supportMode,
      userId,
      orgId,
      orgSlug,
      planId,
      isHighComplianceProject,
      includesLogsSnippets,
      isExplorerEnabled: explorerEnabled,
      requestedModel,
      effectiveModel,
      systemProviderOptions,
      abortSignal: abortController.signal,
      timeout: { totalMs: Math.max(0, ASSISTANT_TIMEOUT_MS - (Date.now() - requestStartedAt)) },
      onSpanCreated: (spanId) => {
        res.setHeader('x-braintrust-span-id', spanId)
      },
    })

    const stream = toUIMessageStream({
      stream: result.stream,
      sendReasoning: true,
      onError: (error) => {
        console.error('Assistant stream error:', error)

        const encoded = encodeNotebookToolError(error)
        if (encoded !== null) return encoded

        if (error == null) {
          return 'unknown error'
        }

        if (typeof error === 'string') {
          return error
        }

        if (error instanceof Error) {
          return error.message
        }

        return JSON.stringify(error)
      },
      // The browser never receives an abort caused by its own disconnect, so any abort that
      // reaches it is the deadline. Chat ignores abort chunks, so flag the message instead.
      messageMetadata: ({ part }): AssistantMessageMetadata =>
        part.type === 'abort' ? { timedOut: true } : undefined,
    })

    // Keep this asynchronous so the TanStack adapter can return the streaming
    // Response immediately. Handle piping failures after headers have been sent.
    // Abort here rather than in toUIMessageStream's onEnd: that callback rebuilds the response
    // message, which fails on approval continuations without the client's original messages.
    void pipeUIMessageStreamToResponse({
      response: res,
      stream,
      headers: { 'Content-Encoding': 'none' },
    })
      .catch((error) => console.error('Error piping Assistant stream:', error))
      // Runs when the stream finishes, aborts, or is cancelled.
      .finally(() => abortController.abort())
  } catch (error) {
    console.error('Error in handlePost:', error)
    if (error instanceof Error) {
      return res.status(500).json({ message: error.message })
    }
    return res.status(500).json({ message: 'An unexpected error occurred.' })
  }
}