Commit Graph

6 Commits

Author SHA1 Message Date
5e07ac040d feat: make backend Docker-ready and web Vercel-ready
- Switch @bytelyst/* deps from link: to private Gitea registry (^0.x)
- Add .npmrc pointing to gitea.bytelyst.com private npm registry
- Rewrite backend/Dockerfile: monorepo root context, pnpm workspace,
  correct EXPOSE 4018, CMD node dist/backend/src/bootstrap.js
- Move vercel.json to repo root with pnpm filter build commands
- Remove web/Dockerfile and web/nginx.conf (web is Vercel-only)
- Remove web service from docker-compose.yml (backend Docker only)
- Document GITEA_NPM_TOKEN requirement in .env.example
- Fix start script path: dist/backend/src/bootstrap.js (rootDir: "..")

PREREQUISITE: Set GITEA_NPM_TOKEN and run pnpm install to regenerate
pnpm-lock.yaml before first Docker build.

Vercel settings: Root Directory = repo root, add GITEA_NPM_TOKEN env var.
Docker build: GITEA_NPM_TOKEN=<token> docker compose build

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-05 19:05:35 -07:00
aaa516122e feat(backend): wire Azure Key Vault secret resolution at startup
- Add bootstrap.ts as new entry point — resolves Key Vault secrets via
  DefaultAzureCredential before config/index.ts is evaluated, so all
  process.env reads pick up KV values (Azure CLI in dev, Managed Identity
  in prod). Falls back to .env if AZURE_KEYVAULT_URL is not set.
- Define INVTTRDG_SECRETS mappings for Cosmos, Azure OpenAI, product-id
- Add AZURE_OPENAI_ENDPOINT / KEY / DEPLOYMENT to config
- aiClient: prefer AzureOpenAIProvider (AI Foundry) when Azure OpenAI
  config is present; falls back to direct OpenAI if not configured
- Add @azure/identity, @azure/keyvault-secrets, @bytelyst/config deps
- Update dev/start scripts to use bootstrap.ts entry point
- Document AZURE_KEYVAULT_URL and Azure OpenAI vars in .env.example

Key Vault: https://kv-mywisprai.vault.azure.net/
Secrets prefix: invttrdg-*

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-05 18:28:47 -07:00
c3651f5696 feat(backend): replace custom aiClient with @bytelyst/llm platform package
- Import PerplexityProvider, OpenAIProvider, GeminiProvider from @bytelyst/llm
- Use createFallbackChain() instead of manual axios fallback loop
- Remove axios and @types/axios — no longer needed
- Preserve AIClient class interface (generateAnalysis, getProviderHealth) —
  no changes required in apiServer.ts or AIAnalysisRule.ts
- Fallback order still driven by config.AI.FALLBACK_LIST

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-05 12:50:34 -07:00
d01ed51bff fix: harden repo verification scripts 2026-04-04 14:35:01 -07:00
d78aeeffc2 feat: adopt platform auth and cosmos trading control 2026-04-04 13:13:08 -07:00
3cbbd6ccaa feat: scaffold trading monorepo foundation 2026-04-04 11:18:21 -07:00