Merge pull request #129 from snexus/feature/bump-version

This PR updates package dependencies to uv‑based management, removes support for llama‑cpp, and introduces an MCP server for semantic search and RAG answer operations.

Dependency upgrades and configuration updates
Removal of llama‑cpp support and introduction of MCP server
Update of configuration and cache handling throughout the codebase

Author: snexus

.flake8

@@ -0,0 +1,2 @@
+[flake8]
+max-line-length = 119

.github/workflows/release.yml

@@ -8,7 +8,7 @@ jobs:
       name: Build source distribution
       runs-on: ubuntu-latest
       steps:
-        - uses: actions/checkout@v3
+        - uses: actions/checkout@v4
 
         - uses: actions/[email protected]
           with:
@@ -22,7 +22,7 @@ jobs:
         - name: Run build
           run: python -m build
 
-        - uses: actions/upload-artifact@v3
+        - uses: actions/upload-artifact@v4
           with:
             path: ./dist/*
 
@@ -34,7 +34,7 @@ jobs:
     runs-on: ubuntu-latest
 
     steps:
-      - uses: actions/download-artifact@v3
+      - uses: actions/download-artifact@v4
         with:
           name: artifact
           path: ./dist

README.md

@@ -4,10 +4,13 @@
 
 [Documentation](https://llm-search.readthedocs.io/en/latest/)
 
-The purpose of this package is to offer a convenient question-answering (RAG) system with a simple YAML-based configuration that enables interaction with multiple collections of local documents. Special attention is given to improvements in various components of the system **in addition to basic LLM-based RAGs** - better document parsing, hybrid search, HyDE enabled search, chat history, deep linking, re-ranking, the ability to customize embeddings, and more. The package is designed to work with custom Large Language Models (LLMs) – whether from OpenAI or installed locally.
+The purpose of this package is to offer an advanced question-answering (RAG) system with a simple YAML-based configuration that enables interaction with a collection of local documents. Special attention is given to improvements in various components of the system **in addition to basic LLM-based RAGs** - better document parsing, hybrid search, HyDE, chat history, deep linking, re-ranking, the ability to customize embeddings, and more. The package is designed to work with custom Large Language Models (LLMs) – whether from OpenAI or installed locally.
+
+Interaction with the package is supported through the built-in frontend, or by exposing an MCP server, allowing clients like Cursor, Windsurf or VSCode GH Copilot to interact with the RAG system.
 
 ## Features
 
+* Fast, incremental parsing and embedding of medium size document bases (tested on up to few gigabytes of markdown and pdfs)
 * Supported document formats
     * Build-in parsers:
         * `.md` - Divides files based on logical components such as headings, subheadings, and code blocks. Supports additional features like cleaning image links, adding custom metadata, and more.
@@ -17,12 +20,13 @@ The purpose of this package is to offer a convenient question-answering (RAG) sy
         * List of formats see [here](https://unstructured-io.github.io/unstructured/core/partition.html).
 
 * Allows interaction with embedded documents, internally supporting the following models and methods (including locally hosted):
-    * OpenAI models (ChatGPT 3.5/4 and Azure OpenAI).
+    * OpenAI compatible models and APIs.
     * HuggingFace models.
-    * Llama cpp supported models - for full list see [here](https://github.com/ggerganov/llama.cpp#description).
 
 * Interoperability with LiteLLM + Ollama via OpenAI API, supporting hundreds of different models (see [Model configuration for LiteLLM](sample_templates/llm/litellm.yaml))
 
+* SSE MCP Server enabling interface with popular MCP clients.
+
 * Generates dense embeddings from a folder of documents and stores them in a vector database ([ChromaDB](https://github.com/chroma-core/chroma)).
   * The following embedding models are supported:
     * Hugging Face embeddings.
@@ -50,12 +54,11 @@ The purpose of this package is to offer a convenient question-answering (RAG) sy
 
 * Supprts optional chat history with question contextualization
 
-
 * Other features
-    * Simple CLI and web interfaces.
+    * Simple web interfaces.
     * Deep linking into document sections - jump to an individual PDF page or a header in a markdown file.
     * Ability to save responses to an offline database for future analysis.
-    * Experimental API
+    * FastAPI based API + MCP server, allo
 
 
 ## Demo

docker/Dockerfile

@@ -8,6 +8,7 @@ Prerequisites
     * Tested with CUDA 11.8 to 12.4 - https://developer.nvidia.com/cuda-toolkit
 * To interact with OpenAI models, create `.env` in the root directory of the repository, containing OpenAI API key. A template for the `.env` file is provided in `.env_template`
 * For parsing `.epub` documents, Pandoc is required - https://pandoc.org/installing.html
+* `uv` - https://github.com/astral-sh/uv#installation
 
 
 
@@ -16,21 +17,16 @@ Install Latest Version
 
 .. code-block:: bash
     
-
     # Create a new environment
-    python3 -m venv .venv 
+    uv venv
 
     # Activate new environment
     source .venv/bin/activate
 
-    # Install packages using pip
-    pip install pyllmsearch
-
     # Optional dependencues for Azure parser
-    pip install "pyllmsearch[azureparser]"
+    uv pip install "pyllmsearch[azureparser]"
 
     # Preferred method (much faster) - install packages using uv
-    pip install uv
     uv pip install pyllmsearch
 
 
@@ -45,20 +41,16 @@ Install from source
 
     git clone https://github.com/snexus/llm-search.git
     cd llm-search
+    # Create a new environment
+    uv venv
+    # Activate new environment
+    source .venv/bin/activate
+    # Install packages using uv
 
-    # Optional - Set variables for llama-cpp to compile with CUDA.
-    # Assuming Nvidia CUDA Toolkit is installed and pointing to `usr/local/cuda` on Ubuntu
-
-    source ./setvars.sh 
-
-    # Optional - Install newest stable torch for CUDA 11.x
-    # pip3 install torch torchvision --index-url https://download.pytorch.org/whl/cu118
-
-    # or for CUDA 12.x version
-    # pip3 install torch torchvision
+    uv sync 
 
-    # Install the package
-    pip install . # or `pip install -e .` for development
+    # Optional - install in the development mode
+    uv pip install -e . # or `pip install -e .` for development
     
     # For Azure parser, install with optional dependencies
-    pip install ."[azureparser]"
+    uv pip install ."[azureparser]"

docker/entrypoint.sh

@@ -46,11 +46,14 @@ To interact with the documents using one of the supported LLMs, follow these ste
 Here `path/to/config/folder` points to a folder of one or more document config files. The tool will scans the configs and allows to switch between them.
 
 
-API (Experimental)
------------------
+API and MCP Server
+------------------
 
-To launch an api, supply a path config file in the `FASTAPI_LLM_CONFIG` environment variable and launch `llmsearchapi` 
+To launch FastAPI/MCP server, supply a path semantic search config file in the `FASTAPI_RAG_CONFIG` and path to llm config in `FASTAPI_LLM_CONFIG` environment variable and launch `llmsearchapi` 
 
 .. code-block:: bash
 
-    FASTAPI_LLM_CONFIG="/path/to/config.yaml" llmsearchapi
+    FASTAPI_RAG_CONFIG="/path/to/config.yaml" FASTAPI_LLM_CONFIG="/path/to/llm.yaml" llmsearchapi
+
+1. The API server will be available at `http://localhost:8000/docs` and can be used to interact with the documents using the LLMs.
+2. The MCP server will be available at `http://localhost:8000/mcp` and can be configured via any MCP client, assuming SSE MCP server which should point to the same URL.

docs/installation.rst

@@ -5,17 +5,43 @@ build-backend = "setuptools.build_meta"
 [project]
 name = "pyllmsearch"
 description = "LLM Powered Advanced RAG Application"
-dynamic = ["dependencies", "version"]
+# dynamic = ["dependencies", "version"]
+dynamic = ["version"]
 keywords = ["llm", "rag", "retrieval-augemented-generation","large-language-models",  "local", "splade", "hyde", "reranking", "chroma", "openai"]
 readme = "README.md"
-requires-python = ">=3.9"
+requires-python = ">=3.10"
 classifiers = [
   "License :: OSI Approved :: MIT License",
   "Operating System :: OS Independent",
   "Programming Language :: Python :: 3 :: Only",
   "Programming Language :: Python :: 3.10",
   "Programming Language :: Python :: 3.11",
 ]
+dependencies = [
+    "langchain-community>=0.3.22",
+    "langchain>=0.3.24",
+    "langchain-huggingface>=0.1.2",
+    "langchain-chroma>=0.2.3",
+    "python-dotenv>=1.1.0",
+    "loguru>=0.7.3",
+    "click>=8.1.8",
+    "openai>=1.76.0",
+    "streamlit>=1.44.1",
+    "tenacity>=9.1.2",
+    "tqdm>=4.67.1",
+    "gmft==0.2.1",
+    "pypdf2>=3.0.1",
+    "pydantic>=2.11.3",
+    "instructorembedding>=1.0.1",
+    "unstructured>=0.17.2",
+    "tiktoken>=0.9.0",
+    "tokenizers>=0.21.1",
+    "langchain-openai>=0.3.14",
+    "python-docx>=1.1.2",
+    "pymupdf>=1.25.5",
+    "termcolor>=3.0.1",
+    "fastapi-mcp>=0.3.3",
+]
 
 [project.optional-dependencies]
 
@@ -35,6 +61,10 @@ azureparser = [
 "azure-identity==1.17.1"
 ]
 
+googleparser = [
+    "google-generativeai>=0.8.5",
+  ]
+
 [project.urls]
 Homepage = "https://github.com/snexus/llm-search"
 Documentation = "https://llm-search.readthedocs.io/en/latest/"
@@ -46,9 +76,30 @@ local_scheme = "no-local-version"
 [tool.setuptools.packages.find]
 where = ["src"]
 
-[tool.setuptools.dynamic]
-dependencies = {file = ["requirements.txt"]}
+# [tool.setuptools.dynamic]
+# dependencies = {file = ["requirements.txt"]}
 
+[tool.flake8]
+docstring-convention = "all"
+ignore = [
+  "D107",
+  "D212",
+  "E501",
+  "W503",
+  "W605",
+  "D203",
+  "D100",
+  "D400",
+  "D415",
+  "D104",
+  "D203",
+  "D213",
+  "D401",
+  "D406",
+  "D417",
+]
+exclude = [ "venv" ]
+max-line-length = 119
 
 [tool.ruff]
 # Decrease the maximum line length to 79 characters.