feat: Phase 2 — embeddings + hybrid memory.search
Adds a small embedder sidecar (Xenova/bge-small-en-v1.5, ONNX, CPU-only) that the web app calls inline on memory.write and memory.update, and on demand from the new memory.search tool. memory.search performs three candidate fetches in parallel — pgvector cosine similarity, Postgres full-text via plainto_tsquery + ts_rank_cd, and tag-set overlap — then fuses them with Reciprocal Rank Fusion (k=60). Each result carries its per-source rank so the model can see *why* a memory surfaced. The migrator boot step gained an idempotent embedding backfill: any row with embedding IS NULL is batched (32 at a time) through the embedder after SQL migrations apply. Safe to run on every boot. New tool memory.update fixes the missing edit path; centralises the re-embed-on-content-change rule alongside write. Stack additions: - apps/embedder/ — Fastify server, persistent /data/models volume so the ~30 MB model only downloads once - apps/web/lib/embedder.ts — typed HTTP client with batched embed + health probe - packages/schemas — MemoryUpdateInput, MemorySearchInput - docker-compose — embedder service, healthcheck, app + migrator both depend_on it healthy; EMBEDDER_URL promoted to a required env var Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
@@ -0,0 +1,62 @@
|
||||
# syntax=docker/dockerfile:1.7
|
||||
# -----------------------------------------------------------------------------
|
||||
# Embedder sidecar.
|
||||
#
|
||||
# Builds from the repo root: docker build -f apps/embedder/Dockerfile .
|
||||
# -----------------------------------------------------------------------------
|
||||
|
||||
FROM node:20-alpine AS base
|
||||
RUN corepack enable
|
||||
WORKDIR /app
|
||||
|
||||
# ---------- deps ----------
|
||||
FROM base AS deps
|
||||
COPY package.json pnpm-workspace.yaml pnpm-lock.yaml .npmrc ./
|
||||
COPY apps/embedder/package.json ./apps/embedder/
|
||||
# Other workspace package.json files needed so pnpm install can resolve the
|
||||
# workspace before --filter narrows things down.
|
||||
COPY apps/web/package.json ./apps/web/
|
||||
COPY packages/schemas/package.json ./packages/schemas/
|
||||
RUN --mount=type=cache,id=pnpm,target=/root/.local/share/pnpm/store \
|
||||
pnpm install --frozen-lockfile --filter @shared-memory/embedder...
|
||||
|
||||
# ---------- builder ----------
|
||||
FROM base AS builder
|
||||
COPY --from=deps /app/node_modules ./node_modules
|
||||
COPY --from=deps /app/apps/embedder/node_modules ./apps/embedder/node_modules
|
||||
COPY . .
|
||||
|
||||
# Compile TS to JS.
|
||||
RUN cd apps/embedder \
|
||||
&& pnpm exec tsc -p tsconfig.json --noEmit false --outDir dist
|
||||
|
||||
# Prune devDependencies so the runtime image only ships production deps.
|
||||
RUN cd apps/embedder \
|
||||
&& pnpm install --prod --frozen-lockfile --filter @shared-memory/embedder...
|
||||
|
||||
# ---------- runner ----------
|
||||
FROM node:20-alpine AS runner
|
||||
WORKDIR /app
|
||||
ENV NODE_ENV=production \
|
||||
PORT=8080 \
|
||||
HOST=0.0.0.0 \
|
||||
MODEL_CACHE_DIR=/data/models
|
||||
|
||||
RUN apk add --no-cache wget \
|
||||
&& addgroup --system --gid 1001 nodejs \
|
||||
&& adduser --system --uid 1001 --ingroup nodejs node-embedder \
|
||||
&& mkdir -p /data/models \
|
||||
&& chown -R node-embedder:nodejs /data
|
||||
|
||||
COPY --from=builder --chown=node-embedder:nodejs /app/apps/embedder/dist ./dist
|
||||
COPY --from=builder --chown=node-embedder:nodejs /app/apps/embedder/node_modules ./node_modules
|
||||
COPY --from=builder --chown=node-embedder:nodejs /app/apps/embedder/package.json ./package.json
|
||||
|
||||
USER node-embedder
|
||||
EXPOSE 8080
|
||||
VOLUME ["/data/models"]
|
||||
|
||||
HEALTHCHECK --interval=15s --timeout=5s --start-period=120s --retries=5 \
|
||||
CMD wget -q -O - http://127.0.0.1:8080/health | grep -q '"ready":true' || exit 1
|
||||
|
||||
CMD ["node", "--enable-source-maps", "dist/index.js"]
|
||||
Reference in New Issue
Block a user