Files
accounted/lib/agent/composer/client.ts
T

70 lines
3.9 KiB
TypeScript

import AnthropicBedrock from '@anthropic-ai/bedrock-sdk'
let cached: AnthropicBedrock | null = null
// Single AnthropicBedrock client for the agent composer + chat loop. Matches
// the credential surface the rest of the codebase already uses (see
// extensions/general/invoice-inbox/lib/extract-invoice-fields.ts) so:
//
// 1. There's no separate ANTHROPIC_API_KEY to provision and rotate.
// 2. All Claude traffic stays in eu-north-1: important for Swedish
// accounting data under BFL retention.
// 3. Failures and quotas show up in one AWS surface, not two.
//
// Trade-off vs. the direct Anthropic API: Bedrock's prompt-cache TTL is
// 5 minutes (default) rather than the 1h the plan §10 specifies. We still
// pass `cache_control: { type: 'ephemeral', ttl: '1h' }` in the system
// prompt assembly: Bedrock currently ignores the explicit TTL and uses 5m.
// Cache effectiveness drops on multi-minute gaps but the loop still works.
// Revisit if/when Bedrock exposes longer TTLs or if cost forces the direct
// API.
export function getAnthropic(): AnthropicBedrock {
if (cached) return cached
// Read Bedrock creds from BEDROCK_AWS_* first, falling back to the plain
// AWS_* names for local dev. On Vercel the functions run on AWS Lambda, whose
// runtime injects its OWN reserved AWS_REGION / AWS_ACCESS_KEY_ID /
// AWS_SECRET_ACCESS_KEY / AWS_SESSION_TOKEN for the platform execution role
// (wrong account, wrong region, no Bedrock access). Those shadow anything set
// in the dashboard, so a hosted deploy MUST use the BEDROCK_AWS_* names or the
// stream comes back empty ("request ended without sending any chunks").
const awsRegion =
process.env.BEDROCK_AWS_REGION || process.env.AWS_REGION || 'eu-north-1'
const awsAccessKey =
process.env.BEDROCK_AWS_ACCESS_KEY_ID || process.env.AWS_ACCESS_KEY_ID
const awsSecretKey =
process.env.BEDROCK_AWS_SECRET_ACCESS_KEY || 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 (instance profile,
// IRSA, EKS pod identity, ...). The two-overload SDK refuses a mix.
cached =
awsAccessKey && awsSecretKey
? new AnthropicBedrock({ awsRegion, awsAccessKey, awsSecretKey })
: new AnthropicBedrock({ awsRegion })
return cached
}
// Bedrock model IDs. Region prefix `eu.` keeps inference inside eu-north-1.
// Both are env-overridable so ops can swap models without a code deploy.
//
// Per plan §14 the composer's atom-selection call should run on Opus 4.7 for
// the higher-stakes selection reasoning. Opus 4.7 is not yet enabled on this
// AWS Bedrock account (403 "not available for this account": request access
// on the AWS console under Bedrock → Model access). For now we point OPUS at
// Sonnet 4.6 so the composer still works; atom selection on Sonnet is still
// good: it's a structured-output call via tool_use forcing, not deep
// reasoning. Flip BEDROCK_OPUS_MODEL_ID back to eu.anthropic.claude-opus-4-7
// once Opus access lands.
export const OPUS_MODEL = process.env.BEDROCK_OPUS_MODEL_ID || 'eu.anthropic.claude-sonnet-4-6'
export const SONNET_MODEL = process.env.BEDROCK_SONNET_MODEL_ID || 'eu.anthropic.claude-sonnet-4-6'
// Extended-thinking budgets (budget_tokens) for the chat intents. These are
// ceilings, not floors: the model spends only what a turn needs, so a generous
// cap improves hard turns (multi-source VAT synthesis, anomaly detection)
// without taxing simple ones. run-turn derives max_tokens = budget + 4096, so
// raising these is safe: no manual max_tokens bookkeeping. Tiered to match the
// model split: DEEP for the Opus / heavy-reasoning intents, STANDARD for the
// rest. Early-stage default favours reasoning quality over token cost; dial
// down here in one place if latency/cost ever bites.
export const THINKING_BUDGET_STANDARD = 6000
export const THINKING_BUDGET_DEEP = 12000