bitflicker64 opened a new pull request, #481:
URL: https://github.com/apache/hugegraph-doc/pull/481
Syncs all five HugeGraph-LLM doc pages (en and cn) with
`apache/hugegraph-ai@main` at version 1.7.0. Every change below traces to a
file on `main`. Paths under `hugegraph-llm/src/hugegraph_llm/` are abbreviated
after the first use.
| Page | Wrong | Changed to | Source |
|---|---|---|---|
| `_index.md` | "HugeGraph-ML is a path dependency rather than a workspace
member" (omits vermeer) | Workspace members are `hugegraph-llm` and
`hugegraph-python-client`; `hugegraph-ml` and `vermeer-python-client` are
editable path deps | `pyproject.toml:87-97` |
| `_index.md` | No repo version stated | Repository version is `1.7.0` |
`pyproject.toml:20` |
| `_index.md` | "HugeGraph-ML and the Python clients: Python 3.10 or later"
| HugeGraph-ML needs >=3.10; the HugeGraph and Vermeer Python clients need
>=3.9 | `hugegraph-ml/pyproject.toml:32`,
`hugegraph-python-client/pyproject.toml:27`,
`vermeer-python-client/pyproject.toml:27` |
| `_index.md`, `hugegraph-llm.md` | "HugeGraph Server 1.5 or later" | 1.3 or
later, 1.5 or later recommended | `hugegraph-llm/README.md:22` |
| `_index.md` | Only the `llm` and `ml` extras mentioned | Table of all
extras: `llm`, `ml`, `python-client`, `vermeer`, `dev`, `nk-llm`, `all`, plus
the module-level `vectordb` | `pyproject.toml:37-54`,
`hugegraph-llm/pyproject.toml:69-73` |
| `_index.md`, `hugegraph-llm.md` | `quick_start.md` was not linked from
anywhere on the site | Added workflow-guide links |
`hugegraph-llm/quick_start.md` |
| `hugegraph-llm.md` | Python requirement stated loosely | `>=3.10,<3.12`
made explicit | `hugegraph-llm/pyproject.toml:30` |
| `hugegraph-llm.md` | No mention of the Compose `.env` bind mount |
Documented the `${PROJECT_PATH}/hugegraph-llm/.env` to
`/home/work/hugegraph-llm/.env` mount and the optional resources mount |
`docker/docker-compose-network.yml:32-36` |
| `hugegraph-llm.md` | No container-image section | Added `hugegraph/rag`
(Dockerfile.llm) and `hugegraph/rag-bin` (Dockerfile.nk, Nuitka), ports,
non-root user, volume, health check | `docker/Dockerfile.llm:37-47`,
`docker/Dockerfile.nk:17-49`, `hugegraph-llm/README.md:69-70` |
| `hugegraph-llm.md` | No mention of the image build script |
`scripts/build_llm_image.sh` builds and tags `hugegraph/graphrag:1.7.0` |
`scripts/build_llm_image.sh:21-29` |
| `hugegraph-llm.md` | Helm chart never mentioned | New Kubernetes section:
`hugegraph/graphrag` image, NodePort 8039 / service 8080 / target 8001, fixed
name `hg-llm-service`, disabled ingress and HPA, config-map instructions, and
the stale `image.tag` default `v0.0.1` |
`docker/charts/hg-llm/values.yaml:9-14,20,34-45,79-85,88-89,128-132` |
| `hugegraph-llm.md` | `HG_DEV_RELOAD` undocumented | Documented as the
uvicorn auto-reload switch | `demo/rag_demo/app.py:208` |
| `hugegraph-llm.md`, `config-reference.md` | `.env` path presented as fixed
at `hugegraph-llm/.env` | Documented the three-step resolution order including
`HUGEGRAPH_LLM_ENV_PATH` | `config/models/base_config.py:28,41-50` |
| `hugegraph-llm.md`, `quick_start.md` | Upload file types never stated |
`.txt`, `.docx`, `.pdf`, multi-select; encrypted and OCR-less PDFs rejected |
`utils/vector_index_utils.py:33-56,66-81` |
| `hugegraph-llm.md`, `quick_start.md` | Graph Schema Generator and Graph
Extraction Prompt Generator missing | Both documented |
`demo/rag_demo/vector_graph_block.py:269-295,413-433` |
| `hugegraph-llm.md`, `quick_start.md` | `Graph Extraction Split Type`
control missing | Documented (`document` / `paragraph` / `sentence`) |
`demo/rag_demo/vector_graph_block.py:403-407` |
| `hugegraph-llm.md`, `quick_start.md` | Batch back-testing panel missing |
Documented `.xlsx` / `.csv` upload, `Max Lines To Show`, template download,
downloadable answers | `demo/rag_demo/rag_block.py:440-463` |
| `hugegraph-llm.md`, `quick_start.md` | Text2Gremlin example-index build
missing | `.json` / `.csv` pairs file, default
`resources/demo/text2gremlin.csv` |
`demo/rag_demo/text2gremlin_block.py:90,107-110,247-249` |
| `hugegraph-llm.md`, `quick_start.md` | "Graph Tools runs graph operations
directly. Admin Tools provides functions such as log access." | Named the
actual actions: Gremlin query, manual backup, beta demo-data init; admin log
tail with 60s refresh, clear and refresh, insecure-token rejection |
`demo/rag_demo/other_block.py:31-50`,
`demo/rag_demo/admin_block.py:73-107,131-146` |
| `hugegraph-llm.md`, `quick_start.md` | Background jobs never mentioned |
Daily 01:00 backup cron and the vid-embedding refresh task |
`demo/rag_demo/other_block.py:53-72` |
| `hugegraph-llm.md`, `config-reference.md` | Vector backend selection had
no config key | `CUR_VECTOR_INDEX` = `Faiss` / `Milvus` / `Qdrant`, default
`Faiss`, also exposed in UI panel 5 | `config/index_config.py:36`,
`utils/vector_index_utils.py:118-140`, `demo/rag_demo/configs_block.py:710-717`
|
| `hugegraph-llm.md` | No programmatic API documented | Added the
`SchedulerSingleton.schedule_flow` example, all 13 flow names, and
`schedule_stream_flow` | `flows/scheduler.py:106,143,188-198`,
`flows/__init__.py:21-34`, `hugegraph-llm/README.md:186-251` |
| `hugegraph-llm.md` | Dev checks used `./style/code_format_and_analysis.sh`
(runs `black`, which is not in the `dev` extra) and bare `pytest` | Replaced
with `uv sync --extra llm --extra dev`, `uv run ruff format --check .` and
`ruff check .`, and the two `SKIP_EXTERNAL_SERVICES=true` pytest splits; added
pre-commit | `style/code_format_and_analysis.sh:44-56`, `pyproject.toml:42-51`,
`hugegraph-llm/AGENTS.md:32,38-43`, `README.md:26-33` |
| `quick_start.md` | Configuration panel not described | Added section 0
listing the five accordions and their apply behaviour |
`demo/rag_demo/configs_block.py:319,353,588,659,710` |
| `quick_start.md` | "A value less than or equal to zero supplies no
templates" | `< 0` skips Text2Gremlin entirely and goes to the predefined
traversal; `0` is zero-shot; `> 0` uses templates, clamped to 0-10 |
`nodes/hugegraph_node/graph_query_node.py:146-149,441-445`,
`nodes/index_node/gremlin_example_index_query.py:41-47`,
`demo/rag_demo/rag_block.py:306-309` |
| `quick_start.md` | `Load into GraphDB` and `Update Vid Embedding`
relationship unclear | `Load into GraphDB (2)` chains `update_vid_embedding`
automatically | `demo/rag_demo/vector_graph_block.py:521-523` |
| `quick_start.md` | Text2Gremlin refer-example count and output fields
missing | 0-10 slider, default 2; four output fields (templated and raw
Gremlin, and their executions) | `demo/rag_demo/text2gremlin_block.py:264-283` |
| `quick_start.md`, `rest-api.md` | Login described only as "the UI and APIs
require USER_TOKEN" | Gradio basic auth uses the fixed user name `rag` with
`USER_TOKEN` as the password; the API uses Bearer and returns 401 with
`WWW-Authenticate` | `demo/rag_demo/app.py:53-62,165-191` |
| `rest-api.md` | No endpoint overview | Table of all 9 POST routes with
their success status codes | `api/rag_api.py:136,179,235,242,265,280,297`,
`api/admin_api.py:53`, `api/graph_extract_api.py:67` |
| `rest-api.md` | Empty-query behaviour undocumented | `/rag`, `/rag/graph`
and `/text2gremlin` return 400 on an empty query; `/rag/graph` returns 400 on
TypeError and 500 otherwise |
`api/rag_api.py:140-144,184-188,223-233,302-306,320-327` |
| `rest-api.md` | `graph_ratio`, `rerank_method`, `near_neighbor_first`
listed without defaults | Defaults `0.5`, `bleu`, `false` added; prompt fields
named individually | `api/models/rag_requests.py:44-46,64-73` |
| `rest-api.md` | `/rag/graph` response keys incomplete | Added `query` and
`graph_result_flag`; noted that `get_vertex_only` swaps `match_vids` for full
vertex details | `api/rag_api.py:204-220` |
| `rest-api.md` | `/graph/extract` fields `example_prompt`, `extract_type`,
`include_meta` missing; no schema validation rules | Full request-field table
plus the inline-schema validation rules |
`api/models/graph_extract_requests.py:34-104` |
| `rest-api.md` | `/graph/extract` `client_config` shape unstated | Only
`graph`, `user`, `pwd`, `gs`; extra fields forbidden; no `url` field |
`api/models/graph_extract_requests.py:25-31` |
| `rest-api.md` | "A successful response always contains ... `meta`" |
`status` is always `succeeded`; `meta` stays empty unless `include_meta` is
true | `api/models/graph_extract_responses.py:23-27`,
`api/graph_extract_api.py:47-55` |
| `rest-api.md` | `/text2gremlin` `example_num` default unstated | Default
`0`, clamped 0-10; missing prompt placeholders fail validation and are named |
`api/models/rag_requests.py:146,161-169`,
`nodes/index_node/gremlin_example_index_query.py:41-45` |
| `rest-api.md` | `/config/graph` field defaults unstated | `user` and `pwd`
default to empty strings, `gs` is optional | `api/models/rag_requests.py:30-35`
|
| `rest-api.md` | `/config/llm` scope unstated | It sets the chat, extract
and text2gql types to the same value | `api/rag_api.py:246-248` |
| `rest-api.md` | Config endpoints described without status codes or
rollback behaviour | All four return 201; llm, embedding and rerank roll back
on exception, `/config/graph` does not | `api/rag_api.py:235-295` |
| `rest-api.md` | `client_config` override behaviour vague | Only fields
present in the request are applied | `api/rag_api.py:129-131` |
| `rest-api.md` | `/logs` described only as "must be a file name under
`logs/`" | Added the `log_file` default, the full rejection rules, the two
distinct 403 paths, and that the response is a `text/plain` `tail -f` stream
replaying the last 125 lines | `api/admin_api.py:30-49,55-70`,
`api/models/rag_requests.py:131-133`, `demo/rag_demo/admin_block.py:28-42` |
| `config-reference.md` | `generate --update` behaviour unstated |
`--update` defaults to true; writes HugeGraph, admin, LLM and index settings
plus the prompt YAML; prompts before overwriting an existing `.env` |
`config/generate.py:31-38`, `config/models/base_config.py:63-82` |
| `config-reference.md` | Loading semantics unstated | Unknown keys ignored,
empty values fall back to defaults, keys case-insensitive |
`config/models/base_config.py:56-61` |
| `config-reference.md` | Config-definition list omitted the index and base
config | Added `config/index_config.py` and `config/models/base_config.py` |
`config/index_config.py`, `config/models/base_config.py` |
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
