feat: redesign as context-guru — provider-agnostic context-engineering

.github/workflows/ci.yaml

@@ -31,7 +31,7 @@ jobs:
     steps:
       - uses: actions/checkout@v4
       - name: Trivy filesystem scan
-        uses: aquasecurity/trivy-action@0.28.0
+        uses: aquasecurity/trivy-action@v0.36.0
         with:
           scan-type: fs
           ignore-unfixed: true

.gitignore

@@ -168,3 +168,6 @@ cython_debug/
 *.test
 *.out
 coverage.txt
+
+# benchmark run output (results summarized in docs/PR, not tracked)
+deploy/eval-containers/sweep-results.csv

CLAUDE.md

@@ -1,26 +1,29 @@
 # CLAUDE.md
 
-Guidance for working in `lab-context-engineering`, a Kagenti platform component.
+Guidance for working in `context-guru` (repo dir `lab-context-engineering`), a Kagenti
+platform component.
 
 ## What this repo is
 
-A single **Go** core that reduces the token cost of LLM agent traffic. It ships as a
-standalone proxy binary (`cmd/proxy`), an importable library (`engine`, `surfaces`), and
-eval-containers wiring. Its lineage is the Python `winnow` prototype (`../winnow`), which
-is the behavioral reference — port its *logic*, re-implement its transport in Go.
+A single **Go** core (`components`) that reduces the token cost of LLM agent traffic,
+operating on bifrost's provider-agnostic chat schema. It ships as a proxy binary
+(`cmd/context-guru-proxy`), an importable library (`components`, `apply`, `schema`,
+`config`, `expand`, `store`), a bifrost `LLMPlugin` adapter (`adapters/bifrost`), and
+eval-containers wiring. Its lineage is the Python `winnow` prototype, the behavioral
+reference — port its *logic*, re-implement its transport in Go.
 
 ## Hard boundaries
 
 - **No AuthBridge / kagenti-extensions code lives here.** That plugin is built in
-  `kagenti-extensions` and depends on this repo. Keep the public API (`engine`,
-  `surfaces`, `config`) clean and importable; never reach into another repo.
-- **Fail open, always.** Any error in any compactor forwards the original request untouched.
-  Reductions must be reversible (markers + rewind store). Never drop content that is only
-  *predicted* unused — `provable_only` is on by default.
+  `kagenti-extensions` and depends on this repo. Keep the public API (`components`,
+  `apply`, `schema`, `config`) clean and importable; never reach into another repo.
+- **Fail open, always.** Any component error/panic reverts that component only; the
+  original request is always forwarded as a valid fallback. Every lossy Offload must be
+  reversible (a `<<cg:HASH>>` marker + the stashed original in the Store).
 
 ## Conventions
 
-- Go 1.25, module `github.com/kagenti/lab-context-engineering`. `make fmt lint test build`.
+- Go 1.26, module `github.com/kagenti/context-guru`. Build needs `CGO_ENABLED=1` (tree-sitter).
 - Match the surrounding code's style; keep packages small and single-purpose.
 - **Commits: DCO sign-off is mandatory** — `git commit -s`. Author as the repo owner.
   AI attribution uses `Assisted-By:` — never `Co-Authored-By`, never a "Generated with"
@@ -29,7 +32,8 @@ is the behavioral reference — port its *logic*, re-implement its transport in
 
 ## Layout
 
-`cmd/proxy` (binary) · `engine` (Transform/Expand + Compactor pipeline) · `surfaces` (wire⇄canonical)
-· `internal/*` (types, extract, relevance, signals, taxonomy, actions, cache, markers,
-rewind, zones, session, compaction, tokens) · `config` · `observability` · `deploy` ·
-`docs`.
+`components` (Component/Reformat/Offload + Pipeline + registry) · `components/{reformat,offload,dsl,all}`
+· `apply` (wire body ⇄ pipeline, byte-lossless splice) · `schema` (bifrost-schema helpers) ·
+`expand` (marker + expand tool loop) · `store` · `session` · `metrics` · `config` ·
+`proxy` · `adapters/bifrost` · `cmd/context-guru-proxy` · `internal/{tokens,treesitter,buildinfo}`
+· `deploy` · `docs`. See [docs/design.md](docs/design.md).

Dockerfile

@@ -1,16 +1,35 @@
-# Build the lab-cx proxy. CGO is required (tree-sitter), so the final image is
-# glibc-based (distroless/base), not static.
-FROM golang:1.25 AS build
+# context-guru-proxy: the LLM-proxy integration and eval-containers gateway.
+#
+# The module uses a local `replace github.com/maximhq/bifrost/core => ../bifrost/core`,
+# so the build context MUST be the PARENT directory that contains both repos:
+#
+#   docker build -f lab-context-engineering/Dockerfile -t context-guru-proxy .
+#
+# (run from .../context-engineering/). CI that builds standalone should either
+# `go mod vendor` first or switch the replace to a pinned published bifrost.
+#
+# CGO is required (tree-sitter for the skeleton component), so the final image is
+# glibc-based. It also carries a shell + curl for the eval-containers
+# start/health scripts under /opt/gateway.
+FROM golang:1.26 AS build
 WORKDIR /src
-COPY . .
+COPY bifrost/ ./bifrost/
+COPY lab-context-engineering/ ./lab-context-engineering/
+WORKDIR /src/lab-context-engineering
 ARG VERSION=dev
 ARG COMMIT=none
 RUN CGO_ENABLED=1 go build \
-	-ldflags "-s -w -X github.com/kagenti/lab-context-engineering/internal/buildinfo.Version=${VERSION} -X github.com/kagenti/lab-context-engineering/internal/buildinfo.Commit=${COMMIT}" \
-	-o /out/lab-cx ./cmd/proxy
+	-ldflags "-s -w -X github.com/kagenti/context-guru/internal/buildinfo.Version=${VERSION} -X github.com/kagenti/context-guru/internal/buildinfo.Commit=${COMMIT}" \
+	-o /out/context-guru-proxy ./cmd/context-guru-proxy
 
-FROM gcr.io/distroless/base-debian12:nonroot
-COPY --from=build /out/lab-cx /usr/local/bin/lab-cx
-EXPOSE 8080
-ENTRYPOINT ["/usr/local/bin/lab-cx"]
-CMD ["proxy"]
+FROM debian:bookworm-slim
+RUN apt-get update && apt-get install -y --no-install-recommends ca-certificates curl \
+	&& rm -rf /var/lib/apt/lists/*
+COPY --from=build /out/context-guru-proxy /opt/gateway/main
+COPY lab-context-engineering/deploy/eval-containers/start /opt/gateway/start
+COPY lab-context-engineering/deploy/eval-containers/health /opt/gateway/health
+RUN chmod +x /opt/gateway/start /opt/gateway/health
+EXPOSE 4000
+# Default entrypoint is the eval-containers gateway wrapper; override with
+# `/opt/gateway/main` to run the proxy directly with flags.
+ENTRYPOINT ["/opt/gateway/start"]

Makefile

@@ -1,5 +1,5 @@
-BINARY := lab-cx
-PKG := github.com/kagenti/lab-context-engineering
+BINARY := context-guru-proxy
+PKG := github.com/kagenti/context-guru
 VERSION ?= $(shell git describe --tags --always --dirty 2>/dev/null || echo dev)
 COMMIT ?= $(shell git rev-parse --short HEAD 2>/dev/null || echo none)
 LDFLAGS := -s -w \
@@ -18,9 +18,9 @@ help: ## Display this help
 		awk 'BEGIN {FS = ":.*?## "}; {printf "  \033[36m%-12s\033[0m %s\n", $$1, $$2}'
 
 .PHONY: build
-build: ## Build the lab-cx binary into ./bin
+build: ## Build the context-guru-proxy binary into ./bin
 	@mkdir -p bin
-	go build -ldflags "$(LDFLAGS)" -o bin/$(BINARY) ./cmd/proxy
+	go build -ldflags "$(LDFLAGS)" -o bin/$(BINARY) ./cmd/context-guru-proxy
 
 .PHONY: test
 test: ## Run all tests with the race detector