Files
accounted/lib/ai/provider.ts
T
Jakob Wennberg c7a75d069d feat(ai): job-shaped AI service with OpenAI-compatible backend, extraction-first; stop extracting every inbox document twice (#1740)
* feat(ai): job-shaped AI service with OpenAI-compatible backend, extraction-first; stop extracting every inbox document twice

Sovereign plan WS1 PR1 (#1406 Tier 2, extraction-first, aligned with the
AI surface audit).

lib/ai grows a job-shaped service (generateText / generateStructured /
extractFromDocument; no streaming members yet, see plan rule R3):
- services/anthropic-family delegates to the existing createAiClient()
  and sends the exact request literals the inbox extractor sent before
  (request-shape tests deep-equal them), so hosted Bedrock stays
  byte-identical.
- services/openai-compatible talks to any chat-completions endpoint
  (BYO Swedish provider) via Vercel AI SDK 6.x, exact-pinned and
  guarded: images as parts, PDFs rasterized with poppler (AI_PDF_MODE)
  or sent natively, AI_VISION / AI_STRICT_JSON declared, honest skips
  (ai_no_vision, pdf_rasterizer_missing) instead of fake failures.
- config.ts: AI_PROVIDER/AI_BASE_URL/AI_API_KEY/AI_MODEL and per-tier
  AI_*_MODEL with the legacy BEDROCK_* names kept as the same overrides;
  getAiStatus() is the single source of truth for "is AI wired up".
- provider.ts: openai-compatible in the auto-detect chain (after Bedrock
  and the direct API); createAiClient() refuses it loudly.

Document extraction moves onto the service and gets the audit's fixes:
- Inbox documents were extracted TWICE (pipeline A ran inside
  uploadDocument() before the inbox row existed, so its dedupe branch
  never fired; 3 707 + 1 666 calls / 30 d). The inbox now declares
  extractionOwner on the upload, the extension yields, and the inbox
  mirrors its single outcome onto document_attachments from every
  writer (sync, deferred, attach, retry, MCP).
- Every "no extraction will ever happen" outcome is stamped
  (skipped:no_ai_entitlement / ai_unconfigured / system_generated /
  ...); the status route maps the quiet ones to 'disabled' on the first
  poll instead of a 30 s client timeout. Prod showed 309 of the 327
  never-extracted uploads were the paywall working silently.
- Self-generated documents (our own invoice PDFs, payout files) are no
  longer OCR'd.
- Agent invoke answers 503 ai_unconfigured when the deployment has no
  assistant backend, distinct from the paywall.

Guard: new direct-ai-client antipattern check (shrink-only allowlist of
the pre-abstraction SDK callers) plus exact pins for @anthropic-ai/sdk,
ai and @ai-sdk/openai-compatible.

Verified: 15 958 unit tests green, guards, lint ratchet, typecheck, and a
live smoke against hosted Bedrock through the new service (ping, streamed
tool turn, thinking+cache, PDF extraction).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

* feat(ai): make AI_API_KEY optional for OpenAI-compatible endpoints (keyless local model servers)

A local model server (llama.cpp's server, Ollama /v1, LM Studio, vLLM)
usually has no auth. Before, the OpenAI-compatible backend required both
AI_BASE_URL and AI_API_KEY to count as configured, so running Accounted on a
local model meant setting a meaningless placeholder key.

- resolveAiProvider / hasAiCredentials: a base URL alone is now enough.
- services/openai-compatible: only send Authorization: Bearer when AI_API_KEY
  is set, so a keyless server is never handed an empty bearer; a hosted
  provider that needs a key still sets it.
- Docs (SELF-HOSTING Option 3: local-model example, key marked optional),
  DECISIONS.

Verified: with no AI_API_KEY, just AI_BASE_URL + AI_MODEL, getAiStatus()
reports configured=true / provider=openai-compatible (live). lib/ai suite
71 green; tsc, guards, lint clean. Bedrock/Anthropic logic unchanged.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

---------

Co-authored-by: Jakob Wennberg <311770904+jakobwennberg-oss@users.noreply.github.com>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
2026-08-20 19:39:08 +02:00

163 lines
7.4 KiB
TypeScript

import Anthropic from '@anthropic-ai/sdk'
import AnthropicBedrock from '@anthropic-ai/bedrock-sdk'
/**
* Which backend AI traffic goes to.
*
* Hosted runs on AWS Bedrock: keeping inference inside eu-north-1 is a
* deliberate BFL/GDPR posture for Swedish accounting data, not an
* implementation detail. Self-hosted deployments generally have no AWS
* account at all, so they get the direct Anthropic API with a plain
* ANTHROPIC_API_KEY, or any OpenAI-compatible endpoint (the Swedish inference
* providers a sovereign self-host points at, or a local model) via
* AI_BASE_URL (+ AI_API_KEY when the endpoint requires auth).
*
* The two Anthropic-family backends share the `messages.create/stream`
* surface this factory hands out. The OpenAI-compatible backend does not: it
* is reachable only through the job-shaped service in lib/ai (getAiService),
* and createAiClient() refuses it loudly rather than returning a client that
* would fail at call time.
*
* See https://github.com/erp-mafia/accounted/issues/1406.
*/
export type AiProvider = 'bedrock' | 'anthropic' | 'openai-compatible'
/**
* Thrown by createAiClient() when the resolved backend has no Anthropic
* messages surface. Surfaces that still call the SDK directly (the chat loop,
* composer, receipt hunt, WhatsApp interpreter) are unavailable on such a
* backend until they move onto getAiService(); routes check
* getAiStatus().assistantAvailable first so users get a 503 instead of this.
*/
export class AiProviderUnsupportedError extends Error {
readonly provider: AiProvider
constructor(provider: AiProvider) {
super(
`AI provider "${provider}" has no Anthropic messages surface; use getAiService() from @/lib/ai instead of createAiClient()`
)
this.name = 'AiProviderUnsupportedError'
this.provider = provider
}
}
export type AiClient = Anthropic | AnthropicBedrock
/**
* Resolve the provider from the environment, or null when neither backend has
* usable credentials.
*
* Precedence is deliberate:
*
* 1. `AI_PROVIDER` wins when set. The escape hatch for a deployment that has
* both credential sets and needs to say which one it means.
* 2. Static AWS keys mean Bedrock. This is what keeps hosted byte-identical:
* an operator who adds an Anthropic key for a side experiment must not
* silently move production inference out of eu-north-1.
* 3. Otherwise an Anthropic key means the direct API. This is the
* self-hosted path.
* 3b. Otherwise an OpenAI-compatible base URL + key means that endpoint.
* This is the sovereign self-hosted path (BYO Swedish provider).
* 4. Otherwise Bedrock without static keys, so the AWS credential provider
* chain (instance profile, IRSA, EKS pod identity) still resolves on
* hosted infrastructure that injects credentials rather than setting env
* vars. `hasAiCredentials()` reports false here: we cannot see the chain
* from this side, so callers that need a cheap pre-flight treat it as
* unconfigured rather than paying a request to find out.
*/
export function resolveAiProvider(): AiProvider {
const explicit = (process.env.AI_PROVIDER ?? '').trim().toLowerCase()
if (explicit === 'bedrock' || explicit === 'anthropic' || explicit === 'openai-compatible') {
return explicit
}
if (process.env.AWS_ACCESS_KEY_ID && process.env.AWS_SECRET_ACCESS_KEY) return 'bedrock'
if (process.env.ANTHROPIC_API_KEY) return 'anthropic'
if (process.env.AI_BASE_URL) return 'openai-compatible'
return 'bedrock'
}
/**
* Whether this deployment has credentials we can see from the environment.
*
* Used by call sites that must degrade quietly rather than throw: document
* extraction returns an empty result instead of failing an upload. Returns
* false for the AWS provider chain (case 4 above) because it is not visible
* here; that path was already treated the same way before the direct-API
* option existed.
*/
export function hasAiCredentials(): boolean {
const provider = resolveAiProvider()
if (provider === 'anthropic') return !!process.env.ANTHROPIC_API_KEY
// AI_API_KEY is optional: a local OpenAI-compatible server (llama.cpp,
// Ollama, LM Studio, vLLM) usually has no auth, so a base URL alone counts
// as configured. When a hosted Swedish provider needs a key, the operator
// sets AI_API_KEY and the service sends it as a Bearer token.
if (provider === 'openai-compatible') return !!process.env.AI_BASE_URL
return !!(process.env.AWS_ACCESS_KEY_ID && process.env.AWS_SECRET_ACCESS_KEY)
}
/**
* Build a client for the resolved Anthropic-family provider. Both expose the
* same `messages.create` / `messages.stream` surface, which is all the direct
* SDK callers in this codebase use. Throws AiProviderUnsupportedError for the
* OpenAI-compatible backend: see the class doc. New code should not call this
* at all; go through getAiService() from @/lib/ai (antipattern guard
* direct-ai-client enforces that outside lib/ai/services).
*/
export function createAiClient(): AiClient {
const provider = resolveAiProvider()
if (provider === 'openai-compatible') throw new AiProviderUnsupportedError(provider)
if (provider === 'anthropic') {
const apiKey = process.env.ANTHROPIC_API_KEY
// Omit the key when unset so the SDK resolves it itself and fails at call
// time: throwing here would take down every route that merely imports a
// module touching AI.
return apiKey ? new Anthropic({ apiKey }) : new Anthropic()
}
const awsRegion = process.env.AWS_REGION || 'eu-north-1'
const awsAccessKey = process.env.AWS_ACCESS_KEY_ID
const awsSecretKey = process.env.AWS_SECRET_ACCESS_KEY
// When both static keys are present, pass them. Otherwise omit them so the
// SDK falls back to the AWS credential provider chain. The two-overload SDK
// refuses a mix.
return awsAccessKey && awsSecretKey
? new AnthropicBedrock({ awsRegion, awsAccessKey, awsSecretKey })
: new AnthropicBedrock({ awsRegion })
}
/**
* Map a bare Anthropic model id to the form the resolved provider expects.
*
* Bedrock needs the `eu.` inference-profile prefix: a bare
* `anthropic.claude-sonnet-5` is rejected because on-demand throughput
* requires the cross-region profile. The direct API takes the bare id and
* rejects both prefixes.
*
* Ids that already carry a provider prefix pass through untouched, so an
* operator-supplied override in either form keeps working.
*/
export function toProviderModelId(bareModelId: string, provider = resolveAiProvider()): string {
if (provider === 'anthropic' || provider === 'openai-compatible') return bareModelId
if (bareModelId.startsWith('eu.') || bareModelId.startsWith('anthropic.')) return bareModelId
return `eu.anthropic.${bareModelId}`
}
/**
* Non-secret identification of the configured credential, for startup logs.
* Anthropic keys carry a public prefix (`sk-ant-api03` for a standard API key,
* `sk-ant-oat` for an OAuth token); AWS access key ids carry `AKIA` for a
* long-term IAM user key and `ASIA` for an STS/role credential. Never returns
* any part of a secret.
*/
export function aiCredentialPrefix(): string | null {
const provider = resolveAiProvider()
if (provider === 'anthropic') {
return process.env.ANTHROPIC_API_KEY?.slice(0, 12) ?? null
}
// OpenAI-compatible keys have no standard public prefix, so there is no
// non-secret slice to log: identify the endpoint instead.
if (provider === 'openai-compatible') return null
return process.env.AWS_ACCESS_KEY_ID?.slice(0, 4) ?? null
}