https://github.com/unit-mesh/co-unit
CoUnit，一个基于 LLM 的虚拟团队接口人（API），通过向量化文档、知识库、SDK和 API 等，结合 LLM 智能化团队间对接与协作。Merge artificial intelligence seamlessly with team collaboration. Leverage intelligent vectorization to process documents, knowledge bases, SDKs, and APIs, empowering teams to unleash their creativity.
https://github.com/unit-mesh/co-unit
aigc genai genai-poc
Last synced: about 1 month ago
JSON representation
CoUnit，一个基于 LLM 的虚拟团队接口人（API），通过向量化文档、知识库、SDK和 API 等，结合 LLM 智能化团队间对接与协作。Merge artificial intelligence seamlessly with team collaboration. Leverage intelligent vectorization to process documents, knowledge bases, SDKs, and APIs, empowering teams to unleash their creativity.
- Host: GitHub
- URL: https://github.com/unit-mesh/co-unit
- Owner: unit-mesh
- License: mpl-2.0
- Created: 2023-08-10T02:54:22.000Z (almost 2 years ago)
- Default Branch: master
- Last Pushed: 2024-03-29T07:01:23.000Z (about 1 year ago)
- Last Synced: 2025-04-01T07:54:18.803Z (2 months ago)
- Topics: aigc, genai, genai-poc
- Language: Rust
- Homepage:
- Size: 15.3 MB
- Stars: 30
- Watchers: 4
- Forks: 3
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- Changelog: CHANGELOG.md
- License: LICENSE
Awesome Lists containing this project
README
CoUnit
> CoUnit，一个基于 LLM 的虚拟团队接口人（API），通过向量化文档、知识库、SDK和 API 等，结合 LLM 智能化团队间对接与协作。
todos:
- Semantic search for Team API
- [x] ArchGuard API：Code, DatabaseMap, HTTP API
- [x] Query for OpenAPI
- [ ] Documents
- [ ] Markdown
- [ ] PDF
- Transpile / Translate language
- [x] Unique language (aka Domain Language) dictionary
- [x] Transpile for Code, Datamap, API
- Prompt strategy
- [x] [HyDE](https://github.com/texttron/hyde)
- [x] ~~Jieba.rs + StarDict spike~~
- [x] Small LLM spike
CoUnit Prompt Strategy
## Uses
1. Download CoUnit-Server binary from [https://github.com/unit-mesh/co-unit/releases](https://github.com/unit-mesh/co-unit/releases)
2. push you `domain language` or config under `domain` directory (support, `.csv` and `.json` format)
3. Run CoUnit-Server
Domain Language:
| native | english | abbreviation | description |
|--------|----------------------|--------------|--------------------------------------------------------------------------|
| CoUnit | collaboration unit | CU | CoUnit，一个基于 LLM 的虚拟团队接口人（API），通过向量化文档、知识库、SDK和 API 等，结合 LLM 智能化团队间对接与协作。 |
| 大语言模型 | Large Language Model | LLM | 大语言模型，是指语言模型的参数量超过 10 亿的语言模型。 |
## Use cases
- AutoDev -> code: [counit](https://github.com/unit-mesh/auto-dev/tree/master/src/main/kotlin/cc/unitmesh/devti/counit).
## Development
Tech stacks:
- Rust, a language empowering everyone to build reliable and efficient software.
- Frameworks:
- Axum, Ergonomic and modular web framework built with Tokio, Tower, and Hyper
- Infrastructure:
- Qdrant, Vector similarity search engine
- Ort, ONNX Runtime is a performance-focused complete scoring engine for Open Neural Network Exchange (ONNX) models.
- Tokenizers, Fast State-of-the-Art Tokenizers optimized for Research and Production.
Setup:
1. Install Rust
2. Clone this repo: `git clone https://github.com/unit-mesh/co-unit`
3. install Qdrant by Docker:
```bash
docker pull qdrant/qdrant
docker run -p 6333:6333 -p 6334:6334 \
-e QDRANT__SERVICE__GRPC_PORT="6334" \
qdrant/qdrant
```
4.Run CoUnit-Server.
### API testing
use [counit-server.http](counit-server.http) to test API.
## Integration example with ArchGuard and AutoDev
AutoDev: [https://github.com/unit-mesh/auto-dev](https://github.com/unit-mesh/auto-dev)
ArchGuard: [https://github.com/archguard/archguard](https://github.com/archguard/archguard)
Full processes:
1. Download ArchGuard CLI (scanner_cli-2.0.x-all.jar) from: [https://github.com/archguard/archguard/releases]
2. Run ArchGuard CLI to upload data to Co-Unit:
```bash
Usage: runner [OPTIONS]
scanner cli
Options:
--type [SOURCE_CODE|GIT|DIFF_CHANGES|SCA|RULE|ARCHITECTURE|ESTIMATE|OPENAPI]
--system-id TEXT system id
--server-url TEXT the base url of the archguard api server
--workspace TEXT the workspace directory
--path TEXT the path of target project
--output TEXT http, csv, json, console
--output-dir TEXT output directory
--analyser-spec TEXT Override the analysers via json.
--slot-spec TEXT Override the slot via json.
--language TEXT language: Java, Kotlin, TypeScript, CSharp,
Python, Golang.
--rules TEXT rules: webapi, test, sql
--features TEXT features: apicalls, datamap.
--repo-id TEXT repository id used for git analysing
--branch TEXT repository branch
--started-at INT TIMESTAMP, the start date of the scanned
commit
--since TEXT COMMIT ID, the specific revision of the
baseline
--until TEXT COMMIT ID, the specific revision of the
target
--depth INT INTEGER, the max loop depth
--with-function-code BOOLEAN, whether to include the function
code
-h, --help Show this message and exit
```
For example:
```bash
java -jar scanner_cli-2.0.6-all.jar --language=Kotlin --path=your_path_to_code --server-url=http://localhost:8765 --repo-id="archguard" --with-function-code --output=http --features=apicalls
```
OpenAPI example:
```bash
java -jar scanner_cli-2.0.6-all.jar --language=Kotlin --path=your_swagger_3_file --server-url=http://localhost:8765 --repo-id="payment" --output=http --type=OPENAPI
```
#### ArchGuard APIs:
```http request
### ArchGuard Code datastrcuture
POST http://127.0.0.1:8765/scanner/:systemId/reporting/class-items
### ArchGuard OpenAPI structure
POST http://127.0.0.1:8765/scanner/:systemId/reporting/openapi
### ArchGuard Service Datamap
POST http://127.0.0.1:8765/scanner/:systemId/reporting/container-services
### ArchGuard Datamap
POST http://127.0.0.1:8765/scanner/:systemId/reporting/datamap-relations
```
## License
The Co-Unit index is licensed under the Apache 2.0 license based
on [https://github.com/BloopAI/bloop](https://github.com/BloopAI/bloop) . See `LICENSE`
in [counit-index](./counit-index).
This code is distributed under the MPL 2.0 license. See `LICENSE` in this directory.