feat: merge develop (#123)
* Support hybrid vector retrieval * Enable figures and table reading in Azure DI * Retrieve with multi-modal * Fix mixing up table * Add txt loader * Add Anthropic Chat * Raising error when retrieving help file * Allow same filename for different people if private is True * Allow declaring extra LLM vendors * Show chunks on the File page * Allow elasticsearch to get more docs * Fix Cohere response (#86) * Fix Cohere response * Remove Adobe pdfservice from dependency kotaemon doesn't rely more pdfservice for its core functionality, and pdfservice uses very out-dated dependency that causes conflict. --------- Co-authored-by: trducng <trungduc1992@gmail.com> * Add confidence score (#87) * Save question answering data as a log file * Save the original information besides the rewritten info * Export Cohere relevance score as confidence score * Fix style check * Upgrade the confidence score appearance (#90) * Highlight the relevance score * Round relevance score. Get key from config instead of env * Cohere return all scores * Display relevance score for image * Remove columns and rows in Excel loader which contains all NaN (#91) * remove columns and rows which contains all NaN * back to multiple joiner options * Fix style --------- Co-authored-by: linhnguyen-cinnamon <cinmc0019@CINMC0019-LinhNguyen.local> Co-authored-by: trducng <trungduc1992@gmail.com> * Track retriever state * Bump llama-index version 0.10 * feat/save-azuredi-mhtml-to-markdown (#93) * feat/save-azuredi-mhtml-to-markdown * fix: replace os.path to pathlib change theflow.settings * refactor: base on pre-commit * chore: move the func of saving content markdown above removed_spans --------- Co-authored-by: jacky0218 <jacky0218@github.com> * fix: losing first chunk (#94) * fix: losing first chunk. * fix: update the method of preventing losing chunks --------- Co-authored-by: jacky0218 <jacky0218@github.com> * fix: adding the base64 image in markdown (#95) * feat: more chunk info on UI * fix: error when reindexing files * refactor: allow more information exception trace when using gpt4v * feat: add excel reader that treats each worksheet as a document * Persist loader information when indexing file * feat: allow hiding unneeded setting panels * feat: allow specific timezone when creating conversation * feat: add more confidence score (#96) * Allow a list of rerankers * Export llm reranking score instead of filter with boolean * Get logprobs from LLMs * Rename cohere reranking score * Call 2 rerankers at once * Run QA pipeline for each chunk to get qa_score * Display more relevance scores * Define another LLMScoring instead of editing the original one * Export logprobs instead of probs * Call LLMScoring * Get qa_score only in the final answer * feat: replace text length with token in file list * ui: show index name instead of id in the settings * feat(ai): restrict the vision temperature * fix(ui): remove the misleading message about non-retrieved evidences * feat(ui): show the reasoning name and description in the reasoning setting page * feat(ui): show version on the main windows * feat(ui): show default llm name in the setting page * fix(conf): append the result of doc in llm_scoring (#97) * fix: constraint maximum number of images * feat(ui): allow filter file by name in file list page * Fix exceeding token length error for OpenAI embeddings by chunking then averaging (#99) * Average embeddings in case the text exceeds max size * Add docstring * fix: Allow empty string when calling embedding * fix: update trulens LLM ranking score for retrieval confidence, improve citation (#98) * Round when displaying not by default * Add LLMTrulens reranking model * Use llmtrulensscoring in pipeline * fix: update UI display for trulen score --------- Co-authored-by: taprosoft <tadashi@cinnamon.is> * feat: add question decomposition & few-shot rewrite pipeline (#89) * Create few-shot query-rewriting. Run and display the result in info_panel * Fix style check * Put the functions to separate modules * Add zero-shot question decomposition * Fix fewshot rewriting * Add default few-shot examples * Fix decompose question * Fix importing rewriting pipelines * fix: update decompose logic in fullQA pipeline --------- Co-authored-by: taprosoft <tadashi@cinnamon.is> * fix: add encoding utf-8 when save temporal markdown in vectorIndex (#101) * fix: improve retrieval pipeline and relevant score display (#102) * fix: improve retrieval pipeline by extending first round top_k with multiplier * fix: minor fix * feat: improve UI default settings and add quick switch option for pipeline * fix: improve agent logics (#103) * fix: improve agent progres display * fix: update retrieval logic * fix: UI display * fix: less verbose debug log * feat: add warning message for low confidence * fix: LLM scoring enabled by default * fix: minor update logics * fix: hotfix image citation * feat: update docx loader for handle merged table cells + handle zip file upload (#104) * feat: update docx loader for handle merged table cells * feat: handle zip file * refactor: pre-commit * fix: escape text in download UI * feat: optimize vector store query db (#105) * feat: optimize vector store query db * feat: add file_id to chroma metadatas * feat: remove unnecessary logs and update migrate script * feat: iterate through file index * fix: remove unused code --------- Co-authored-by: taprosoft <tadashi@cinnamon.is> * fix: add openai embedidng exponential back-off * fix: update import download_loader * refactor: codespell * fix: update some default settings * fix: update installation instruction * fix: default chunk length in simple QA * feat: add share converstation feature and enable retrieval history (#108) * feat: add share converstation feature and enable retrieval history * fix: update share conversation UI --------- Co-authored-by: taprosoft <tadashi@cinnamon.is> * fix: allow exponential backoff for failed OCR call (#109) * fix: update default prompt when no retrieval is used * fix: create embedding for long image chunks * fix: add exception handling for additional table retriever * fix: clean conversation & file selection UI * fix: elastic search with empty doc_ids * feat: add thumbnail PDF reader for quick multimodal QA * feat: add thumbnail handling logic in indexing * fix: UI text update * fix: PDF thumb loader page number logic * feat: add quick indexing pipeline and update UI * feat: add conv name suggestion * fix: minor UI change * feat: citation in thread * fix: add conv name suggestion in regen * chore: add assets for usage doc * chore: update usage doc * feat: pdf viewer (#110) * feat: update pdfviewer * feat: update missing files * fix: update rendering logic of infor panel * fix: improve thumbnail retrieval logic * fix: update PDF evidence rendering logic * fix: remove pdfjs built dist * fix: reduce thumbnail evidence count * chore: update gitignore * fix: add js event on chat msg select * fix: update css for viewer * fix: add env var for PDFJS prebuilt * fix: move language setting to reasoning utils --------- Co-authored-by: phv2312 <kat87yb@gmail.com> Co-authored-by: trducng <trungduc1992@gmail.com> * feat: graph rag (#116) * fix: reload server when add/delete index * fix: rework indexing pipeline to be able to disable vectorstore and splitter if needed * feat: add graphRAG index with plot view * fix: update requirement for graphRAG and lighten unnecessary packages * feat: add knowledge network index (#118) * feat: add Knowledge Network index * fix: update reader mode setting for knet * fix: update init knet * fix: update collection name to index pipeline * fix: missing req --------- Co-authored-by: jeff52415 <jeff.yang@cinnamon.is> * fix: update info panel return for graphrag * fix: retriever setting graphrag * feat: local llm settings (#122) * feat: expose context length as reasoning setting to better fit local models * fix: update context length setting for agents * fix: rework threadpool llm call * fix: fix improve indexing logic * fix: fix improve UI * feat: add lancedb * fix: improve lancedb logic * feat: add lancedb vectorstore * fix: lighten requirement * fix: improve lanceDB vs * fix: improve UI * fix: openai retry * fix: update reqs * fix: update launch command * feat: update Dockerfile * feat: add plot history * fix: update default config * fix: remove verbose print * fix: update default setting * fix: update gradio plot return * fix: default gradio tmp * fix: improve lancedb docstore * fix: fix question decompose pipeline * feat: add multimodal reader in UI * fix: udpate docs * fix: update default settings & docker build * fix: update app startup * chore: update documentation * chore: update README * chore: update README --------- Co-authored-by: trducng <trungduc1992@gmail.com> * chore: update README * chore: update README --------- Co-authored-by: trducng <trungduc1992@gmail.com> Co-authored-by: cin-ace <ace@cinnamon.is> Co-authored-by: Linh Nguyen <70562198+linhnguyen-cinnamon@users.noreply.github.com> Co-authored-by: linhnguyen-cinnamon <cinmc0019@CINMC0019-LinhNguyen.local> Co-authored-by: cin-jacky <101088014+jacky0218@users.noreply.github.com> Co-authored-by: jacky0218 <jacky0218@github.com> Co-authored-by: kan_cin <kan@cinnamon.is> Co-authored-by: phv2312 <kat87yb@gmail.com> Co-authored-by: jeff52415 <jeff.yang@cinnamon.is>
This commit is contained in:
Tuan Anh Nguyen Dang (Tadashi_Cin)
committed by
GitHub
GitHub
parent
86d60e1649
commit
2570e11501
206
README.md
206
README.md
@@ -1,12 +1,12 @@
|
||||
# kotaemon
|
||||
|
||||
An open-source tool for chatting with your documents. Built with both end users and
|
||||
An open-source clean & customizable RAG UI for chatting with your documents. Built with both end users and
|
||||
developers in mind.
|
||||
|
||||
https://github.com/Cinnamon/kotaemon/assets/25688648/815ecf68-3a02-4914-a0dd-3f8ec7e75cd9
|
||||
|
||||
|
||||
[Source Code](https://github.com/Cinnamon/kotaemon) |
|
||||
[Live Demo](https://huggingface.co/spaces/cin-model/kotaemon-public)
|
||||
[Live Demo](https://huggingface.co/spaces/taprosoft/kotaemon) |
|
||||
[Source Code](https://github.com/Cinnamon/kotaemon)
|
||||
|
||||
[User Guide](https://cinnamon.github.io/kotaemon/) |
|
||||
[Developer Guide](https://cinnamon.github.io/kotaemon/development/) |
|
||||
@@ -14,20 +14,23 @@ https://github.com/Cinnamon/kotaemon/assets/25688648/815ecf68-3a02-4914-a0dd-3f8
|
||||
|
||||
[](https://www.python.org/downloads/release/python-31013/)
|
||||
[](https://github.com/psf/black)
|
||||
<a href="https://hub.docker.com/r/taprosoft/kotaemon" target="_blank">
|
||||
<img src="https://img.shields.io/badge/docker_pull-kotaemon:v1.0-brightgreen" alt="docker pull taprosoft/kotaemon:v1.0"></a>
|
||||
[](https://codeium.com)
|
||||
|
||||
This project would like to appeal to both end users who want to do QA on their
|
||||
documents and developers who want to build their own QA pipeline.
|
||||
## Introduction
|
||||
|
||||
This project serves as a functional RAG UI for both end users who want to do QA on their
|
||||
documents and developers who want to build their own RAG pipeline.
|
||||
|
||||
- For end users:
|
||||
- A local Question Answering UI for RAG-based QA.
|
||||
- A clean & minimalistic UI for RAG-based QA.
|
||||
- Supports LLM API providers (OpenAI, AzureOpenAI, Cohere, etc) and local LLMs
|
||||
(currently only GGUF format is supported via `llama-cpp-python`).
|
||||
- Easy installation scripts, no environment setup required.
|
||||
(via `ollama` and `llama-cpp-python`).
|
||||
- Easy installation scripts.
|
||||
- For developers:
|
||||
- A framework for building your own RAG-based QA pipeline.
|
||||
- See your RAG pipeline in action with the provided UI (built with Gradio).
|
||||
- Share your pipeline so that others can use it.
|
||||
- A framework for building your own RAG-based document QA pipeline.
|
||||
- Customize and see your RAG pipeline in action with the provided UI (built with Gradio).
|
||||
|
||||
```yml
|
||||
+----------------------------------------------------------------------------+
|
||||
@@ -45,78 +48,128 @@ documents and developers who want to build their own QA pipeline.
|
||||
```
|
||||
|
||||
This repository is under active development. Feedback, issues, and PRs are highly
|
||||
appreciated. Your input is valuable as it helps us persuade our business guys to support
|
||||
open source.
|
||||
appreciated.
|
||||
|
||||
## Key Features
|
||||
|
||||
- **Host your own document QA (RAG) web-UI**. Support multi-user login, organize your files in private / public collections, collaborate and share your favorite chat with others.
|
||||
|
||||
- **Organize your LLM & Embedding models**. Support both local LLMs & popular API providers (OpenAI, Azure, Ollama, Groq).
|
||||
|
||||
- **Hybrid RAG pipeline**. Sane default RAG pipeline with hybrid (full-text & vector) retriever + re-ranking to ensure best retrieval quality.
|
||||
|
||||
- **Multi-modal QA support**. Perform Question Answering on multiple documents with figures & tables support. Support multi-modal document parsing (selectable options on UI).
|
||||
|
||||
- **Advance citations with document preview**. By default the system will provide detailed citations to ensure the correctness of LLM answers. View your citations (incl. relevant score) directly in the _in-browser PDF viewer_ with highlights. Warning when retrieval pipeline return low relevant articles.
|
||||
|
||||
- **Support complex reasoning methods**. Use question decomposition to answer your complex / multi-hop question. Support agent-based reasoning with ReAct, ReWOO and other agents.
|
||||
|
||||
- **Configurable settings UI**. You can adjust most important aspects of retrieval & generation process on the UI (incl. prompts).
|
||||
|
||||
- **Extensible**. Being built on Gradio, you are free to customize / add any UI elements as you like. Also, we aim to support multiple strategies for document indexing & retrieval. `GraphRAG` indexing pipeline is provided as an example.
|
||||
|
||||
|
||||
|
||||
## Installation
|
||||
|
||||
### For end users
|
||||
|
||||
This document is intended for developers. If you just want to install and use the app as
|
||||
it, please follow the [User Guide](https://cinnamon.github.io/kotaemon/).
|
||||
it is, please follow the non-technical [User Guide](https://cinnamon.github.io/kotaemon/) (WIP).
|
||||
|
||||
### For developers
|
||||
|
||||
```shell
|
||||
# Create a environment
|
||||
python -m venv kotaemon-env
|
||||
#### With Docker (recommended)
|
||||
|
||||
# Activate the environment
|
||||
source kotaemon-env/bin/activate
|
||||
- Use this command to launch the server
|
||||
|
||||
# Install the package
|
||||
pip install git+https://github.com/Cinnamon/kotaemon.git
|
||||
```
|
||||
docker run \
|
||||
-e GRADIO_SERVER_NAME=0.0.0.0 \
|
||||
-e GRADIO_SERVER_PORT=7860 \
|
||||
-p 7860:7860 -it --rm \
|
||||
taprosoft/kotaemon:v1.0
|
||||
```
|
||||
|
||||
### For Contributors
|
||||
Navigate to `http://localhost:7860/` to access the web UI.
|
||||
|
||||
#### Without Docker
|
||||
|
||||
- Clone and install required packages on a fresh python environment.
|
||||
|
||||
```shell
|
||||
# Clone the repo
|
||||
git clone git@github.com:Cinnamon/kotaemon.git
|
||||
# optional (setup env)
|
||||
conda create -n kotaemon python=3.10
|
||||
conda activate kotaemon
|
||||
|
||||
# Create a environment
|
||||
python -m venv kotaemon-env
|
||||
|
||||
# Activate the environment
|
||||
source kotaemon-env/bin/activate
|
||||
# clone this repo
|
||||
git clone https://github.com/Cinnamon/kotaemon
|
||||
cd kotaemon
|
||||
|
||||
# Install the package in editable mode
|
||||
pip install -e "libs/kotaemon[all]"
|
||||
pip install -e "libs/ktem"
|
||||
pip install -e "."
|
||||
|
||||
# Setup pre-commit
|
||||
pre-commit install
|
||||
```
|
||||
|
||||
## Creating your application
|
||||
- View and edit your environment variables (API keys, end-points) in `.env`.
|
||||
|
||||
In order to create your own application, you need to prepare these files:
|
||||
- (Optional) To enable in-browser PDF_JS viewer, download [PDF_JS_DIST](https://github.com/mozilla/pdf.js/releases/download/v4.0.379/pdfjs-4.0.379-dist.zip) and extract it to `libs/ktem/ktem/assets/prebuilt`
|
||||
|
||||
<img src="docs/images/pdf-viewer-setup.png" alt="pdf-setup" width="300">
|
||||
|
||||
- Start the web server:
|
||||
|
||||
```shell
|
||||
python app.py
|
||||
```
|
||||
|
||||
The app will be automatically launched in your browser.
|
||||
|
||||
Default username / password are: `admin` / `admin`. You can setup additional users directly on the UI.
|
||||
|
||||
|
||||
|
||||
## Customize your application
|
||||
|
||||
By default, all application data are stored in `./ktem_app_data` folder. You can backup or copy this folder to move your installation to a new machine.
|
||||
|
||||
For advance users or specific use-cases, you can customize those files:
|
||||
|
||||
- `flowsettings.py`
|
||||
- `app.py`
|
||||
- `.env` (Optional)
|
||||
- `.env`
|
||||
|
||||
### `flowsettings.py`
|
||||
|
||||
This file contains the configuration of your application. You can use the example
|
||||
[here](https://github.com/Cinnamon/kotaemon/blob/main/libs/ktem/flowsettings.py) as the
|
||||
[here](flowsettings.py) as the
|
||||
starting point.
|
||||
|
||||
### `app.py`
|
||||
<details>
|
||||
|
||||
This file is where you create your Gradio app object. This can be as simple as:
|
||||
<summary>Notable settings</summary>
|
||||
|
||||
```python
|
||||
from ktem.main import App
|
||||
```
|
||||
# setup your preferred document store (with full-text search capabilities)
|
||||
KH_DOCSTORE=(Elasticsearch | LanceDB | SimpleFileDocumentStore)
|
||||
|
||||
app = App()
|
||||
demo = app.make()
|
||||
demo.launch()
|
||||
# setup your preferred vectorstore (for vector-based search)
|
||||
KH_VECTORSTORE=(ChromaDB | LanceDB
|
||||
|
||||
# Enable / disable multimodal QA
|
||||
KH_REASONINGS_USE_MULTIMODAL=True
|
||||
|
||||
# Setup your new reasoning pipeline or modify existing one.
|
||||
KH_REASONINGS = [
|
||||
"ktem.reasoning.simple.FullQAPipeline",
|
||||
"ktem.reasoning.simple.FullDecomposeQAPipeline",
|
||||
"ktem.reasoning.react.ReactAgentPipeline",
|
||||
"ktem.reasoning.rewoo.RewooAgentPipeline",
|
||||
]
|
||||
)
|
||||
```
|
||||
|
||||
### `.env` (Optional)
|
||||
</details>
|
||||
|
||||
### `.env`
|
||||
|
||||
This file provides another way to configure your models and credentials.
|
||||
|
||||
@@ -159,18 +212,22 @@ AZURE_OPENAI_EMBEDDINGS_DEPLOYMENT=text-embedding-ada-002
|
||||
|
||||
#### Local models
|
||||
|
||||
- Pros:
|
||||
- Privacy. Your documents will be stored and process locally.
|
||||
- Choices. There are a wide range of LLMs in terms of size, domain, language to choose
|
||||
from.
|
||||
- Cost. It's free.
|
||||
- Cons:
|
||||
- Quality. Local models are much smaller and thus have lower generative quality than
|
||||
paid APIs.
|
||||
- Speed. Local models are deployed using your machine so the processing speed is
|
||||
limited by your hardware.
|
||||
##### Using ollama OpenAI compatible server
|
||||
|
||||
##### Find and download a LLM
|
||||
Install [ollama](https://github.com/ollama/ollama) and start the application.
|
||||
|
||||
Pull your model (e.g):
|
||||
|
||||
```
|
||||
ollama pull llama3.1:8b
|
||||
ollama pull nomic-embed-text
|
||||
```
|
||||
|
||||
Set the model names on web UI and make it as default.
|
||||
|
||||
|
||||
|
||||
##### Using GGUF with llama-cpp-python
|
||||
|
||||
You can search and download a LLM to be ran locally from the [Hugging Face
|
||||
Hub](https://huggingface.co/models). Currently, these model formats are supported:
|
||||
@@ -187,33 +244,26 @@ Here are some recommendations and their size in memory:
|
||||
- [Qwen1.5-1.8B-Chat-GGUF](https://huggingface.co/Qwen/Qwen1.5-1.8B-Chat-GGUF/resolve/main/qwen1_5-1_8b-chat-q8_0.gguf?download=true):
|
||||
around 2 GB
|
||||
|
||||
##### Enable local models
|
||||
Add a new LlamaCpp model with the provided model name on the web uI.
|
||||
|
||||
To add a local model to the model pool, set the `LOCAL_MODEL` variable in the `.env`
|
||||
file to the path of the model file.
|
||||
|
||||
```shell
|
||||
LOCAL_MODEL=<full path to your model file>
|
||||
```
|
||||
|
||||
Here is how to get the full path of your model file:
|
||||
|
||||
- On Windows 11: right click the file and select `Copy as Path`.
|
||||
</details>
|
||||
|
||||
## Start your application
|
||||
## Adding your own RAG pipeline
|
||||
|
||||
Simply run the following command:
|
||||
#### Custom reasoning pipeline
|
||||
|
||||
```shell
|
||||
python app.py
|
||||
```
|
||||
First, check the default pipeline implementation in
|
||||
[here](libs/ktem/ktem/reasoning/simple.py). You can make quick adjustment to how the default QA pipeline work.
|
||||
|
||||
The app will be automatically launched in your browser.
|
||||
Next, if you feel comfortable adding new pipeline, add new `.py` implementation in `libs/ktem/ktem/reasoning/` and later include it in `flowssettings` to enable it on the UI.
|
||||
|
||||
|
||||
#### Custom indexing pipeline
|
||||
|
||||
## Customize your application
|
||||
Check sample implementation in `libs/ktem/ktem/index/file/graph`
|
||||
|
||||
(more instruction WIP).
|
||||
|
||||
## Developer guide
|
||||
|
||||
Please refer to the [Developer Guide](https://cinnamon.github.io/kotaemon/development/)
|
||||
for more details.
|
||||
|
||||
Reference in New Issue
Block a user