Skip to content

Route dream synthesis through gateway and add embedding dim migration#1013

Draft
cxbitz wants to merge 1 commit into
garrytan:masterfrom
cxbitz:worktree-gbrain-openai-dream-worker
Draft

Route dream synthesis through gateway and add embedding dim migration#1013
cxbitz wants to merge 1 commit into
garrytan:masterfrom
cxbitz:worktree-gbrain-openai-dream-worker

Conversation

@cxbitz
Copy link
Copy Markdown

@cxbitz cxbitz commented May 15, 2026

Summary

  • Route dream significance/synthesis subagent paths through the existing AI gateway so LiteLLM/OpenAI-compatible routes work without direct Anthropic dependencies.
  • Add Ollama bge-m3 embedding support with model-specific 1024d defaults and provider probes.
  • Add gbrain migrate-embedding-dim to resize existing vector schemas, clear incompatible embeddings/query cache, update config, and re-embed safely.

Verification

  • git diff --check
  • bun run typecheck
  • bun test test/embedding-dim-check.test.ts test/subagent-handler.test.ts test/cycle-synthesize.test.ts test/ai/gateway-chat.test.ts test/ai/gateway.test.ts test/ai/recipes-existing-regression.test.ts test/ai/no-batch-cap-suppression.serial.test.ts (132 pass / 0 fail)
  • Isolated live Ollama flow: 768d init -> migrate to bge-m3 1024d -> dream embed -> query -> stale dry-run clean

Add Ollama bge-m3 embedding support and a migrate-embedding-dim command for existing vector schemas. The migration clears incompatible vectors/query cache, updates config, and supports re-embedding at the new dimension.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant