Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 0 additions & 2 deletions AGENTS.md
Original file line number Diff line number Diff line change
Expand Up @@ -57,7 +57,6 @@ NORTH_STAR.md
- [ ] **本轮启动时已做 Step 1b 对齐检查并显式输出**(覆盖 NORTH_STAR / Phase / ROUNDS / 跳序四问)
- [ ] `mvn17 test -pl graph-rag-core` 全绿
- [ ] 没有跳过的测试(`@Disabled` / `@Ignore`)
- [ ] 没有引入 `com.kuaishou.*` 或其他内部包
- [ ] 如改动了 `GraphDatabaseClient` 接口:Neo4j 实现 + 所有 stub 都已同步
- [ ] 没有删除任何 `Noop*Impl` 类
- [ ] 单类 < 500 行,单方法 < 100 行(超出需在 ROUNDS 里说明理由)
Expand All @@ -74,7 +73,6 @@ NORTH_STAR.md
## AI 禁止事项(违反即视为损坏 SSOT)

- ❌ 跳过测试直接 commit
- ❌ 在 `graph-rag-core` 任意层 import 内部包(`com.kuaishou.*` / 私有 SDK)
- ❌ 修改 `GraphDatabaseClient` 接口而不同步更新所有实现
- ❌ 删除 `Noop*Impl` 实现(它们是测试与 examples 的兜底)
- ❌ 在 `application/` 或 `domain/` 层直接引用 `org.neo4j.driver.*`(必须走接口)
Expand Down
27 changes: 24 additions & 3 deletions CHANGELOG.md
Original file line number Diff line number Diff line change
Expand Up @@ -7,6 +7,28 @@

## [Unreleased]

### Added — Pre-release polish (Round 7 + 7.1)

- **Lightweight hero GIF** (`docs/assets/demo-short.gif`, ~553 KB / ~27 s / 1240×760): `--variant=graph`, real local Ollama, one short DAU question, full graph context + explicit lineage trail rendered. Now sits at the top of README / README.zh.md so first-time visitors on slow networks see a structured graph answer before the longer side-by-side GIF loads. Reproducible via new `docs/assets/demo-short.tape`. The existing 60-second `demo-side-by-side.gif` is preserved as the linked "extended demo".
- **`run-cli.sh` now honors `VARIANT=` env** (defaults to `both`), so the same launcher script is used by both the hero tape (`VARIANT=graph`) and the extended tape (default).
- **Status wording switched from `public alpha` to `public preview`** across README badge / README body / README.zh.md / SECURITY.md / CONTRIBUTING.md. R6's "Public alpha polish" CHANGELOG heading kept as-is (historical record).
- **Bilingual README**: added `README.zh.md` (full 中文 translation, same 13-section structure as English) with language switcher links at the top of both files. README badges now point at the real `MarcelLeon/graph-rag-harness` GitHub Actions CI workflow instead of the `OWNER/REPO` placeholder, and the GIF-recipe wording references the in-repo `./docs/assets/run-cli.sh` launcher (was a stale `/tmp/run-cli.sh` path).
- **Troubleshooting section in README + CONTRIBUTING**: explains why `mvn -pl graph-rag-examples -am spring-boot:run` fails with "Unable to find a suitable main class" (the `-am` includes the root pom module, which has no main class), and gives two safe recipes: (1) split into a one-time `install` for upstream modules + a plain `spring-boot:run` for the example, or (2) just run the pre-built fat jar. Quick-start commands updated accordingly so first-time cloners do not hit this trap.
- **Side-by-side dogfooding GIF** (`docs/assets/demo-side-by-side.gif`, 820 KB / 60 s / 1320×1100): a real local Ollama run (`qwen3:8b` chat + `nomic-embed-text` embedding) that shows the dogfooding CLI answering one DAU question in `--variant=both` mode, with `PLAIN` and `GRAPH` blocks visible in the same frame and the graph answer carrying the explicit `(Graph lineage: DAU --[FORMULA_USES]--> ...)` trail. README front page now embeds this GIF directly.
- **Repo-tracked recording recipe**: `docs/assets/demo.tape` (vhs script, Dracula theme, hides JVM warm-up, deterministic timings) and `docs/assets/run-cli.sh` (port-8088 wrapper that runs `--mode=cli --variant=both` with quiet logs). Anyone can re-record the GIF with `brew install vhs && vhs docs/assets/demo.tape` — the README "Record the demo GIF yourself" section is now a 3-line recipe.
- **GitHub-friendly README rewrite**: added status / CI / license / Java / Spring AI badges, a 60-second terminal preview, an honest "what is solid / what is in progress" summary, and now an embedded demo GIF instead of an ASCII mock-up.
- **`docs/assets/` directory**: now hosts the demo GIF, the vhs tape, and the launcher script that produces it.
- **`docs/playbooks/sample-graph-v2-design.md`** rewritten as a zero-touch, conditionally-go fixture upgrade: locks the design to `graph-rag-examples/` only (no `graph-rag-core` enum / interface changes), maps every new business relation onto existing `STRUCTURAL_TYPES`, defines a 12-node / 14-edge ceiling, adds a regression / rollback / acceptance checklist, and gates the start of v2-mini behind "L3 < +1.0 after Round 7-8 prompt and judge tuning".

### Changed

- **Public eval headline updated**: latest real Ollama run shows L3 plain 2.50 / graph 3.00 (**Δ = +0.50**), overall plain 2.80 / graph 3.00 (Δ = +0.20). README, STATUS, and CHANGELOG aligned. NORTH_STAR milestone of L3 ≥ +1.0 is still tracked honestly as "not yet".
- **Test suite reference numbers refreshed**: `mvn17 test -pl graph-rag-core` now reports 65 tests green (12 + 6 + 4 + 43), up from the 62 figure quoted in earlier docs.

### Notes

- Round 7 itself touched no production code in `graph-rag-core`, `graph-rag-spring-boot-starter`, or `graph-rag-examples`. Round 7.1 added only documentation assets (`docs/assets/*`). The 65-test `graph-rag-core` baseline is unchanged.

### Added — Public alpha polish (Round 6)

- **Interactive dogfooding CLI** (`graph-rag-examples/.../cli/CliRunner.java`): run the example app with `--mode=cli --variant=plain|graph|both` and ask questions directly from the terminal using local Ollama. This reuses the same answer services as the HTTP demo and eval harness.
Expand All @@ -28,7 +50,7 @@

### Notes

- Latest real Ollama eval now shows **early positive L3 uplift** of graph-RAG over plain-RAG (**+0.25** on the latest run, overall **+0.10**), which is enough to support honest public alpha positioning but still **below** the NORTH_STAR milestone of **+1.0** on L3.
- Latest real Ollama eval now shows **positive L3 uplift** of graph-RAG over plain-RAG (**+0.50** on the 2026-05-26 run, overall **+0.20**), which supports honest public alpha positioning but is still **below** the NORTH_STAR milestone of **+1.0** on L3.

### Added — Spring AI demo end-to-end (Round 5)

Expand Down Expand Up @@ -86,5 +108,4 @@
- 20 个单元测试,全绿

### Removed
- 所有内部依赖:KGraph SDK / Cortex / KsBoot
- 内部包名 `com.kuaishou.*`,统一改为 `io.github.graphrag.*`
- 所有内部依赖:KGraph SDK / Cortex / KsBoot
8 changes: 7 additions & 1 deletion CONTRIBUTING.md
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@

Thanks for your interest in contributing to `graph-rag-harness`.

This repository is currently in **public alpha**:
This repository is currently in **public preview**:

- the demo is runnable,
- the core abstractions are real,
Expand Down Expand Up @@ -106,6 +106,12 @@ JAVA_HOME=$JAVA17_HOME mvn -pl graph-rag-examples -am -DskipTests compile -q
If your change touches the example app, also try a manual smoke test:

```bash
# Install upstream modules into the local cache once (or whenever they change).
JAVA_HOME=$JAVA17_HOME mvn -pl graph-rag-core,graph-rag-spring-boot-starter -am -DskipTests install -q

# Then run the example app on its own — DO NOT add -am here. With -am, Maven
# would try to run spring-boot:run on the root pom module first and fail with
# "Unable to find a suitable main class". See README "Troubleshooting".
JAVA_HOME=$JAVA17_HOME mvn -pl graph-rag-examples spring-boot:run \
-Dspring-boot.run.arguments="--mode=cli --variant=both --server.port=8086"
```
Expand Down
Loading
Loading