# **ReasonChain**
**ReasonChain** is a modular and extensible AI reasoning library for creating intelligent agents capable of executing advanced reasoning processes. It supports **Chain of Thought (CoT)** reasoning, **Tree of Thought (ToT)** reasoning, **Parallel Chains**, and integrates with **Retrieval-Augmented Generation (RAG)** for enhanced knowledge management.
---
### **Features**
1. **Customizable Reasoning Pipelines**:
- Seamless integration of **Chain of Thought (CoT)**, **Tree of Thought (ToT)**, Parallel Chains, and Hybrid Pipelines.
- Facilitates task decomposition, execution, and validation.
2. **RAG Integration**:
- Retrieve and augment responses using long-term memory stored in vector databases like FAISS.
3. **Short-term and Long-term Memory**:
- Session-based short-term memory for reasoning chains.
- Persistent long-term memory powered by FAISS.
4. **LLM Compatibility**:
- Supports OpenAI GPT, Llama, and other models for robust reasoning and summarization.
5. **Utility Tools**:
- Adaptive complexity evaluation for reasoning tasks.
- Centralized model management for handling LLM interactions.
6. **Supported File Types**:
- Extract text, tables, and figures from a wide range of file formats, including:
- **Documents**: PDF, Word, RTF, Markdown, HTML, LaTeX.
- **Spreadsheets**: Excel, CSV.
- **Multimedia**: Images, Videos, Audio.
- **E-books**: EPUB, MOBI.
- Provides modular extractors tailored for each file type, ensuring efficient data retrieval.
7. **Domain Templates**:
- Pre-built reasoning templates tailored for specific industries and applications:
- **Customer Support**: Automates handling of customer inquiries and escalations.
- **Healthcare**: Assists in diagnosis recommendations and treatment plans.
- **Finance**: Analyzes trends, generates financial reports, and forecasts.
- **Retail**: Provides insights on inventory optimization and customer behavior.
- **Real Estate**: Summarizes market data and assists in property evaluations.
- Fully customizable to meet domain-specific requirements.
- Simplifies the setup of agents for new use cases.
---
## **Installation**
### 1. Install via pip
You can install **ReasonChain** directly from PyPI:
```bash
pip install reasonchain
```
### 2. Clone the Repository (for development)
Alternatively, clone the repository to access examples and editable source code:
```bash
git clone https://github.com/sunnybedi990/reasonchain.git
cd reasonchain
```
### 3. Install Dependencies
```bash
pip install -r requirements.txt
```
### 4. Install as an Editable Package
```bash
pip install -e .
```
### 5. Configure API Keys
Create a `.env` file with your API keys:
```env
# Authentication Keys
LLAMA_CLOUD_API_KEY= # API key for LLama Parser
OPENAI_API_KEY= # API key for OpenAI
GROQ_API_KEY= # API key for Groq
# Vector Database Configuration
VECTOR_DB_PATH=vector_dbs/ # Path to store vector databases
USE_GPU=true # Set to "true" for GPU usage or "false" for CPU
# Pinecone Configuration
PINECONE_API_KEY= # API key for Pinecone
# Weaviate Configuration
WEAVIATE_API_KEY= # API key for Weaviate
WEAVIATE_CLUSTER_URL= # Cluster URL for Weaviate
# Qdrant Configuration
QDRANT_API_KEY= # API key for Qdrant (optional)
QDRANT_ADMIN_API_KEY= # Admin API key for Qdrant (optional)
QDRANT_CLUSTER_URL= # Cluster URL for Qdrant (optional)
```
---
## **Usage**
### 1. Initialize an Agent
```python
from reasonchain import Agent
agent = Agent(name="ResearchBot", model_name="gpt-4", api="openai")
```
### 2. Build a CoT Reasoning Pipeline
```python
from reasonchain.cot_pipeline import CoTPipeline
pipeline = CoTPipeline(agent)
pipeline.add_step("Understand the user's query.")
pipeline.add_step("Retrieve data from the knowledge base.")
pipeline.add_step("Generate a detailed response.")
```
### 3. Execute the Pipeline
```python
response = pipeline.execute(agent.model_manager)
print(response)
```
### 1. Build and Integrate Agents with RAG
ReasonChain makes it easy to create multi-agent systems and integrate them with RAG for context-enhanced reasoning.
```python
from reasonchain.agent import Agent, MultiAgentSystem
from reasonchain.rag.vector.add_to_vector_db import add_pdf_to_vector_db
from reasonchain.rag.rag_main import query_vector_db
from reasonchain.utils import (
store_in_shared_memory,
retrieve_from_shared_memory,
collaborate_on_task,
assign_and_execute_task
)
# Initialize the Multi-Agent System
multi_agent_system = MultiAgentSystem()
# Create and register agents
agent1 = Agent(name="AgentAlpha", role="extractor", model_name="gpt-4o", api="openai")
agent2 = Agent(name="AgentBeta", role="analyst", model_name="llama3.1:latest", api="ollama")
agent3 = Agent(name="AgentGamma", role="summarizer", model_name="llama-3.1-8b-instant", api="groq")
multi_agent_system.register_agent(agent1)
multi_agent_system.register_agent(agent2)
multi_agent_system.register_agent(agent3)
# Add a document to the vector database
add_pdf_to_vector_db(
pdf_path="path/to/tesla-report.pdf",
db_path="vector_db.index",
db_type="faiss",
embedding_provider="sentence_transformers",
embedding_model="all-mpnet-base-v2",
use_gpu=True
)
# Query the vector database
query = "Extract financial highlights from the Tesla Q-10 report."
response = query_vector_db(
db_path="vector_db.index",
db_type="faiss",
query=query,
embedding_provider="sentence_transformers",
embedding_model="all-mpnet-base-v2"
)
# Store and retrieve data in shared memory
store_in_shared_memory(agent1.shared_memory, "financial_highlights", response)
highlights = retrieve_from_shared_memory(agent2.shared_memory, "financial_highlights")
# Assign tasks and collaborate
task_description = "Analyze trends and summarize Tesla's financial highlights."
collaborators = ["AgentBeta", "AgentGamma"]
successful_agents = collaborate_on_task(multi_agent_system, collaborators, task_description)
print(f"Successful Agents: {successful_agents}")
```
---
## **Examples**
Explore more scripts in the `examples/` directory:
- **`rag_pipeline_example.py`**: Example of using RAG for context-enhanced reasoning.
- **`multi-agent_collaboration.py`**: Multi-agent collaboration example.
- **`tree_of_thought_example.py`**: Demonstrates Tree of Thought reasoning.
- **`hybrid_reasoning_example.py`**: Combines multiple reasoning methods.
---
## **Supported File Types**
ReasonChain supports extracting text, tables, and figures from the following file types. It uses specialized extractors for each type to ensure accurate and comprehensive data retrieval.
### **Supported Extensions and Extractors**
| File Type | Extensions | Extractor Library/Tools | Extracted Data |
|-------------------------|------------------------|----------------------------------------------------------|--------------------------|
| **PDF Documents** | `.pdf` | PyMuPDF, Camelot, PDFPlumber, LlamaParse | Text, Tables, Figures |
| **Word Documents** | `.doc`, `.docx` | python-docx | Text, Tables, Figures |
| **Excel Files** | `.xls`, `.xlsx` | pandas | Tables |
| **CSV Files** | `.csv` | pandas | Tables |
| **HTML Files** | `.html`, `.htm` | BeautifulSoup, pandas | Text, Tables, Figures |
| **Markdown Files** | `.md` | re, pandas | Text, Tables, Figures |
| **Rich Text Files** | `.rtf` | pyrtf | Text |
| **E-Books** | `.epub`, `.mobi` | ebooklib, BeautifulSoup | Text |
| **Images** | `.png`, `.jpg`, `.jpeg`, `.tiff` | pytesseract (OCR) | Text |
| **Presentation Files** | `.ppt`, `.pptx` | python-pptx | Text, Figures |
| **Audio Files** | `.mp3`, `.wav` | SpeechRecognition, Whisper | Text |
| **Video Files** | `.mp4`, `.avi`, `.mov`| moviepy, pytesseract (OCR) | Text, Figures |
| **LaTeX Files** | `.tex` | regex, plain text processing | Text |
---
### **Feature Highlights for Extractors**
1. **PDF Extraction**:
- Handles structured text, tables, and embedded images.
- LlamaParse integration for multimodal capabilities.
2. **Word and Presentation Files**:
- Extracts text from paragraphs, tables, and embedded images.
- Saves embedded figures locally for further processing.
3. **E-books and Markdown**:
- Processes text and embedded images or hyperlinks.
- Parses tables represented as plain text in Markdown.
4. **Images and Videos**:
- Extracts frames from videos and applies OCR for textual content.
- Processes scanned documents and infographics using pytesseract.
5. **Audio Extraction**:
- Converts audio to text using SpeechRecognition or Whisper.
6. **Rich Text and LaTeX**:
- Converts RTF files into plain text.
- Removes LaTeX commands to provide clean text content.
---
### Domain Templates
ReasonChain provides **pre-built reasoning templates** tailored for specific industries and applications. These templates simplify the process of creating intelligent agents by embedding domain-specific logic and workflows. They can be used as-is or customized to meet unique requirements.
#### Available Domain Templates
| **Domain** | **Description** |
|--------------------------|---------------------------------------------------------------------------------|
| **Customer Support** | Automates resolution workflows, ticket prioritization, and customer query handling. |
| **E-commerce** | Assists in product recommendations, inventory optimization, and user behavior analysis. |
| **Education** | Supports personalized learning plans, content recommendations, and assessments. |
| **Energy** | Optimizes energy consumption, monitors grid performance, and supports sustainability planning. |
| **Finance** | Analyzes financial trends, generates reports, and forecasts revenue and expenses. |
| **Gaming & Entertainment** | Provides real-time analytics for player engagement and content recommendations. |
| **Healthcare** | Assists in diagnosis, treatment recommendations, and patient care management. |
| **HR & Recruitment** | Streamlines candidate screening, skill assessments, and onboarding workflows. |
| **Legal** | Summarizes legal documents, case analysis, and compliance checks. |
| **Marketing** | Automates campaign analysis, audience targeting, and content generation. |
| **Real Estate** | Evaluates property data, summarizes market trends, and supports investment decisions. |
| **Retail** | Enhances customer experience, tracks sales data, and suggests optimization strategies. |
| **Supply Chain** | Monitors logistics, optimizes inventory, and forecasts demand. |
| **Transportation** | Provides route optimization, fleet management, and traffic pattern analysis. |
| **Travel & Hospitality** | Personalizes travel recommendations, optimizes bookings, and improves guest experiences. |
#### Example: Using the Customer Support Template
```python
from reasonchain.domain_templates.customer_support import CustomerSupportTemplate
# Initialize a Customer Support reasoning pipeline
pipeline = CustomerSupportTemplate()
pipeline.add_step("Analyze the customer's query.")
pipeline.add_step("Retrieve similar case resolutions.")
pipeline.add_step("Provide a detailed response to resolve the issue.")
response = pipeline.execute()
print(response)
```
#### Customizing Templates
Each template is fully customizable. You can modify the reasoning steps, add additional logic, or integrate with external APIs to extend the capabilities.
For example, to add a step in the Finance template:
```python
from reasonchain.domain_templates.finance import FinanceTemplate
pipeline = FinanceTemplate()
pipeline.add_step("Analyze the quarterly revenue data.")
pipeline.add_step("Provide insights into expenditure trends.")
pipeline.add_step("Forecast revenue for the next quarter.")
response = pipeline.execute()
print(response)
```
#### Extendability
- Easily create new templates for other domains by extending the base template classes.
- Integrate templates with Retrieval-Augmented Generation (RAG) for enhanced reasoning and data retrieval.
---
## Project Structure
The ReasonChain library is organized into the following core components:
- **Core Library (`reasonchain/`)**:
- Contains the main modules for reasoning pipelines, RAG integration, domain templates, and utilities.
- **Examples (`examples/`)**:
- Demonstrates the use of RAG pipelines, multi-agent systems, domain-specific templates, and hybrid reasoning.
- **Unit Tests (`tests/`)**:
- Validates the functionality of various components.
- **Pre-trained Models and Outputs**:
- Includes directories like `fine_tuned_model/`, `markdown_images/`, and `parsed_chunks/` for storing models and processed outputs.
For a detailed breakdown of the project structure, see the [CONTRIBUTING.md](CONTRIBUTING.md) or the `docs/` folder.
---
## **Key Features**
1. **Dynamic Pipelines**:
- Choose from CoT, ToT, Parallel Chains, or Hybrid Pipelines.
2. **Knowledge Integration**:
- Augment reasoning with RAG for external data retrieval.
3. **Scalable Memory**:
- Manage short-term and long-term memory effectively.
4. **LLM Compatibility**:
- Seamlessly work with OpenAI GPT, Llama, and similar models.
---
## **Future Enhancements**
1. **Domain-Specific Templates**:
- Add pre-trained reasoning templates for specialized applications.
2. **Agent Collaboration**:
- Enable seamless teamwork between agents.
3. **Extended RAG Support**:
- Integrate with Pinecone, Milvus, and more vector databases.
4. **Fine-Tuning Support**:
- Incorporate custom fine-tuned models for advanced reasoning.
---
## **Contributing**
Contributions are welcome! Follow these steps:
1. Fork the repository.
2. Create a new branch:
```bash
git checkout -b feature/your-feature
```
3. Commit your changes:
```bash
git commit -m "Add your feature"
```
4. Push to your fork:
```bash
git push origin feature/your-feature
```
5. Open a pull request.
---
## **License**
This project is licensed under the MIT License. See the `LICENSE` file for more details.
Raw data
{
"_id": null,
"home_page": "https://github.com/sunnybedi990/reasonchain",
"name": "reasonchain",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.7",
"maintainer_email": null,
"keywords": null,
"author": "Baljindersingh Bedi",
"author_email": "baljindersinghbedi409@gmail.com",
"download_url": "https://files.pythonhosted.org/packages/08/f4/a77dd2d1c896147968be8096c70d3697cc36d8ec6d63c6542daa00845f37/reasonchain-0.2.16.tar.gz",
"platform": null,
"description": "# **ReasonChain**\n\n**ReasonChain** is a modular and extensible AI reasoning library for creating intelligent agents capable of executing advanced reasoning processes. It supports **Chain of Thought (CoT)** reasoning, **Tree of Thought (ToT)** reasoning, **Parallel Chains**, and integrates with **Retrieval-Augmented Generation (RAG)** for enhanced knowledge management.\n\n---\n\n### **Features**\n\n1. **Customizable Reasoning Pipelines**:\n - Seamless integration of **Chain of Thought (CoT)**, **Tree of Thought (ToT)**, Parallel Chains, and Hybrid Pipelines.\n - Facilitates task decomposition, execution, and validation.\n\n2. **RAG Integration**:\n - Retrieve and augment responses using long-term memory stored in vector databases like FAISS.\n\n3. **Short-term and Long-term Memory**:\n - Session-based short-term memory for reasoning chains.\n - Persistent long-term memory powered by FAISS.\n\n4. **LLM Compatibility**:\n - Supports OpenAI GPT, Llama, and other models for robust reasoning and summarization.\n\n5. **Utility Tools**:\n - Adaptive complexity evaluation for reasoning tasks.\n - Centralized model management for handling LLM interactions.\n\n6. **Supported File Types**:\n - Extract text, tables, and figures from a wide range of file formats, including:\n - **Documents**: PDF, Word, RTF, Markdown, HTML, LaTeX.\n - **Spreadsheets**: Excel, CSV.\n - **Multimedia**: Images, Videos, Audio.\n - **E-books**: EPUB, MOBI.\n - Provides modular extractors tailored for each file type, ensuring efficient data retrieval.\n\n7. **Domain Templates**:\n - Pre-built reasoning templates tailored for specific industries and applications:\n - **Customer Support**: Automates handling of customer inquiries and escalations.\n - **Healthcare**: Assists in diagnosis recommendations and treatment plans.\n - **Finance**: Analyzes trends, generates financial reports, and forecasts.\n - **Retail**: Provides insights on inventory optimization and customer behavior.\n - **Real Estate**: Summarizes market data and assists in property evaluations.\n - Fully customizable to meet domain-specific requirements.\n - Simplifies the setup of agents for new use cases.\n\n---\n\n## **Installation**\n\n### 1. Install via pip\nYou can install **ReasonChain** directly from PyPI:\n```bash\npip install reasonchain\n```\n\n### 2. Clone the Repository (for development)\nAlternatively, clone the repository to access examples and editable source code:\n```bash\ngit clone https://github.com/sunnybedi990/reasonchain.git\ncd reasonchain\n```\n\n### 3. Install Dependencies\n```bash\npip install -r requirements.txt\n```\n\n### 4. Install as an Editable Package\n```bash\npip install -e .\n```\n\n### 5. Configure API Keys\nCreate a `.env` file with your API keys:\n```env\n# Authentication Keys\nLLAMA_CLOUD_API_KEY= # API key for LLama Parser\nOPENAI_API_KEY= # API key for OpenAI\nGROQ_API_KEY= # API key for Groq\n\n# Vector Database Configuration\nVECTOR_DB_PATH=vector_dbs/ # Path to store vector databases\nUSE_GPU=true # Set to \"true\" for GPU usage or \"false\" for CPU\n\n# Pinecone Configuration\nPINECONE_API_KEY= # API key for Pinecone\n\n# Weaviate Configuration\nWEAVIATE_API_KEY= # API key for Weaviate\nWEAVIATE_CLUSTER_URL= # Cluster URL for Weaviate\n\n# Qdrant Configuration\nQDRANT_API_KEY= # API key for Qdrant (optional)\nQDRANT_ADMIN_API_KEY= # Admin API key for Qdrant (optional)\nQDRANT_CLUSTER_URL= # Cluster URL for Qdrant (optional)\n```\n\n---\n\n## **Usage**\n\n### 1. Initialize an Agent\n```python\nfrom reasonchain import Agent\n\nagent = Agent(name=\"ResearchBot\", model_name=\"gpt-4\", api=\"openai\")\n```\n\n### 2. Build a CoT Reasoning Pipeline\n```python\nfrom reasonchain.cot_pipeline import CoTPipeline\n\npipeline = CoTPipeline(agent)\npipeline.add_step(\"Understand the user's query.\")\npipeline.add_step(\"Retrieve data from the knowledge base.\")\npipeline.add_step(\"Generate a detailed response.\")\n```\n\n### 3. Execute the Pipeline\n```python\nresponse = pipeline.execute(agent.model_manager)\nprint(response)\n```\n\n### 1. Build and Integrate Agents with RAG\nReasonChain makes it easy to create multi-agent systems and integrate them with RAG for context-enhanced reasoning.\n\n```python\nfrom reasonchain.agent import Agent, MultiAgentSystem\nfrom reasonchain.rag.vector.add_to_vector_db import add_pdf_to_vector_db\nfrom reasonchain.rag.rag_main import query_vector_db\nfrom reasonchain.utils import (\n store_in_shared_memory,\n retrieve_from_shared_memory,\n collaborate_on_task,\n assign_and_execute_task\n)\n\n# Initialize the Multi-Agent System\nmulti_agent_system = MultiAgentSystem()\n\n# Create and register agents\nagent1 = Agent(name=\"AgentAlpha\", role=\"extractor\", model_name=\"gpt-4o\", api=\"openai\")\nagent2 = Agent(name=\"AgentBeta\", role=\"analyst\", model_name=\"llama3.1:latest\", api=\"ollama\")\nagent3 = Agent(name=\"AgentGamma\", role=\"summarizer\", model_name=\"llama-3.1-8b-instant\", api=\"groq\")\n\nmulti_agent_system.register_agent(agent1)\nmulti_agent_system.register_agent(agent2)\nmulti_agent_system.register_agent(agent3)\n\n# Add a document to the vector database\nadd_pdf_to_vector_db(\n pdf_path=\"path/to/tesla-report.pdf\",\n db_path=\"vector_db.index\",\n db_type=\"faiss\",\n embedding_provider=\"sentence_transformers\",\n embedding_model=\"all-mpnet-base-v2\",\n use_gpu=True\n)\n\n# Query the vector database\nquery = \"Extract financial highlights from the Tesla Q-10 report.\"\nresponse = query_vector_db(\n db_path=\"vector_db.index\",\n db_type=\"faiss\",\n query=query,\n embedding_provider=\"sentence_transformers\",\n embedding_model=\"all-mpnet-base-v2\"\n)\n\n# Store and retrieve data in shared memory\nstore_in_shared_memory(agent1.shared_memory, \"financial_highlights\", response)\nhighlights = retrieve_from_shared_memory(agent2.shared_memory, \"financial_highlights\")\n\n# Assign tasks and collaborate\ntask_description = \"Analyze trends and summarize Tesla's financial highlights.\"\ncollaborators = [\"AgentBeta\", \"AgentGamma\"]\nsuccessful_agents = collaborate_on_task(multi_agent_system, collaborators, task_description)\n\nprint(f\"Successful Agents: {successful_agents}\")\n```\n\n---\n\n## **Examples**\n\nExplore more scripts in the `examples/` directory:\n- **`rag_pipeline_example.py`**: Example of using RAG for context-enhanced reasoning.\n- **`multi-agent_collaboration.py`**: Multi-agent collaboration example.\n- **`tree_of_thought_example.py`**: Demonstrates Tree of Thought reasoning.\n- **`hybrid_reasoning_example.py`**: Combines multiple reasoning methods.\n\n---\n\n## **Supported File Types**\n\nReasonChain supports extracting text, tables, and figures from the following file types. It uses specialized extractors for each type to ensure accurate and comprehensive data retrieval.\n\n### **Supported Extensions and Extractors**\n| File Type | Extensions | Extractor Library/Tools | Extracted Data |\n|-------------------------|------------------------|----------------------------------------------------------|--------------------------|\n| **PDF Documents** | `.pdf` | PyMuPDF, Camelot, PDFPlumber, LlamaParse | Text, Tables, Figures |\n| **Word Documents** | `.doc`, `.docx` | python-docx | Text, Tables, Figures |\n| **Excel Files** | `.xls`, `.xlsx` | pandas | Tables |\n| **CSV Files** | `.csv` | pandas | Tables |\n| **HTML Files** | `.html`, `.htm` | BeautifulSoup, pandas | Text, Tables, Figures |\n| **Markdown Files** | `.md` | re, pandas | Text, Tables, Figures |\n| **Rich Text Files** | `.rtf` | pyrtf | Text |\n| **E-Books** | `.epub`, `.mobi` | ebooklib, BeautifulSoup | Text |\n| **Images** | `.png`, `.jpg`, `.jpeg`, `.tiff` | pytesseract (OCR) | Text |\n| **Presentation Files** | `.ppt`, `.pptx` | python-pptx | Text, Figures |\n| **Audio Files** | `.mp3`, `.wav` | SpeechRecognition, Whisper | Text |\n| **Video Files** | `.mp4`, `.avi`, `.mov`| moviepy, pytesseract (OCR) | Text, Figures |\n| **LaTeX Files** | `.tex` | regex, plain text processing | Text |\n\n---\n\n### **Feature Highlights for Extractors**\n\n1. **PDF Extraction**:\n - Handles structured text, tables, and embedded images.\n - LlamaParse integration for multimodal capabilities.\n\n2. **Word and Presentation Files**:\n - Extracts text from paragraphs, tables, and embedded images.\n - Saves embedded figures locally for further processing.\n\n3. **E-books and Markdown**:\n - Processes text and embedded images or hyperlinks.\n - Parses tables represented as plain text in Markdown.\n\n4. **Images and Videos**:\n - Extracts frames from videos and applies OCR for textual content.\n - Processes scanned documents and infographics using pytesseract.\n\n5. **Audio Extraction**:\n - Converts audio to text using SpeechRecognition or Whisper.\n\n6. **Rich Text and LaTeX**:\n - Converts RTF files into plain text.\n - Removes LaTeX commands to provide clean text content.\n\n---\n\n### Domain Templates\n\nReasonChain provides **pre-built reasoning templates** tailored for specific industries and applications. These templates simplify the process of creating intelligent agents by embedding domain-specific logic and workflows. They can be used as-is or customized to meet unique requirements.\n\n#### Available Domain Templates\n\n| **Domain** | **Description** |\n|--------------------------|---------------------------------------------------------------------------------|\n| **Customer Support** | Automates resolution workflows, ticket prioritization, and customer query handling. |\n| **E-commerce** | Assists in product recommendations, inventory optimization, and user behavior analysis. |\n| **Education** | Supports personalized learning plans, content recommendations, and assessments. |\n| **Energy** | Optimizes energy consumption, monitors grid performance, and supports sustainability planning. |\n| **Finance** | Analyzes financial trends, generates reports, and forecasts revenue and expenses. |\n| **Gaming & Entertainment** | Provides real-time analytics for player engagement and content recommendations. |\n| **Healthcare** | Assists in diagnosis, treatment recommendations, and patient care management. |\n| **HR & Recruitment** | Streamlines candidate screening, skill assessments, and onboarding workflows. |\n| **Legal** | Summarizes legal documents, case analysis, and compliance checks. |\n| **Marketing** | Automates campaign analysis, audience targeting, and content generation. |\n| **Real Estate** | Evaluates property data, summarizes market trends, and supports investment decisions. |\n| **Retail** | Enhances customer experience, tracks sales data, and suggests optimization strategies. |\n| **Supply Chain** | Monitors logistics, optimizes inventory, and forecasts demand. |\n| **Transportation** | Provides route optimization, fleet management, and traffic pattern analysis. |\n| **Travel & Hospitality** | Personalizes travel recommendations, optimizes bookings, and improves guest experiences. |\n\n#### Example: Using the Customer Support Template\n\n```python\nfrom reasonchain.domain_templates.customer_support import CustomerSupportTemplate\n\n# Initialize a Customer Support reasoning pipeline\npipeline = CustomerSupportTemplate()\npipeline.add_step(\"Analyze the customer's query.\")\npipeline.add_step(\"Retrieve similar case resolutions.\")\npipeline.add_step(\"Provide a detailed response to resolve the issue.\")\n\nresponse = pipeline.execute()\nprint(response)\n```\n\n#### Customizing Templates\nEach template is fully customizable. You can modify the reasoning steps, add additional logic, or integrate with external APIs to extend the capabilities.\n\nFor example, to add a step in the Finance template:\n```python\nfrom reasonchain.domain_templates.finance import FinanceTemplate\n\npipeline = FinanceTemplate()\npipeline.add_step(\"Analyze the quarterly revenue data.\")\npipeline.add_step(\"Provide insights into expenditure trends.\")\npipeline.add_step(\"Forecast revenue for the next quarter.\")\n\nresponse = pipeline.execute()\nprint(response)\n```\n\n#### Extendability\n- Easily create new templates for other domains by extending the base template classes.\n- Integrate templates with Retrieval-Augmented Generation (RAG) for enhanced reasoning and data retrieval.\n\n---\n\n## Project Structure\n\nThe ReasonChain library is organized into the following core components:\n\n- **Core Library (`reasonchain/`)**:\n - Contains the main modules for reasoning pipelines, RAG integration, domain templates, and utilities.\n- **Examples (`examples/`)**:\n - Demonstrates the use of RAG pipelines, multi-agent systems, domain-specific templates, and hybrid reasoning.\n- **Unit Tests (`tests/`)**:\n - Validates the functionality of various components.\n- **Pre-trained Models and Outputs**:\n - Includes directories like `fine_tuned_model/`, `markdown_images/`, and `parsed_chunks/` for storing models and processed outputs.\n\nFor a detailed breakdown of the project structure, see the [CONTRIBUTING.md](CONTRIBUTING.md) or the `docs/` folder.\n\n---\n\n## **Key Features**\n\n1. **Dynamic Pipelines**:\n - Choose from CoT, ToT, Parallel Chains, or Hybrid Pipelines.\n\n2. **Knowledge Integration**:\n - Augment reasoning with RAG for external data retrieval.\n\n3. **Scalable Memory**:\n - Manage short-term and long-term memory effectively.\n\n4. **LLM Compatibility**:\n - Seamlessly work with OpenAI GPT, Llama, and similar models.\n\n---\n\n## **Future Enhancements**\n\n1. **Domain-Specific Templates**:\n - Add pre-trained reasoning templates for specialized applications.\n\n2. **Agent Collaboration**:\n - Enable seamless teamwork between agents.\n\n3. **Extended RAG Support**:\n - Integrate with Pinecone, Milvus, and more vector databases.\n\n4. **Fine-Tuning Support**:\n - Incorporate custom fine-tuned models for advanced reasoning.\n\n---\n\n## **Contributing**\n\nContributions are welcome! Follow these steps:\n1. Fork the repository.\n2. Create a new branch:\n ```bash\n git checkout -b feature/your-feature\n ```\n3. Commit your changes:\n ```bash\n git commit -m \"Add your feature\"\n ```\n4. Push to your fork:\n ```bash\n git push origin feature/your-feature\n ```\n5. Open a pull request.\n\n---\n\n## **License**\n\nThis project is licensed under the MIT License. See the `LICENSE` file for more details.\n",
"bugtrack_url": null,
"license": "MIT",
"summary": "A modular AI reasoning library for building intelligent agents.",
"version": "0.2.16",
"project_urls": {
"Homepage": "https://github.com/sunnybedi990/reasonchain"
},
"split_keywords": [],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "5eb84eab9b02bea11317525142262d92574d8655fd2956a833f852fc2feb7454",
"md5": "ce8f386cdbb6f2d4f6549244ceb1abda",
"sha256": "8c6d2f59721064d51622695fb4e3d0061e69198355c3e518c28c631eef510189"
},
"downloads": -1,
"filename": "reasonchain-0.2.16-py3-none-any.whl",
"has_sig": false,
"md5_digest": "ce8f386cdbb6f2d4f6549244ceb1abda",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.7",
"size": 7026471,
"upload_time": "2025-01-04T20:34:16",
"upload_time_iso_8601": "2025-01-04T20:34:16.971084Z",
"url": "https://files.pythonhosted.org/packages/5e/b8/4eab9b02bea11317525142262d92574d8655fd2956a833f852fc2feb7454/reasonchain-0.2.16-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "08f4a77dd2d1c896147968be8096c70d3697cc36d8ec6d63c6542daa00845f37",
"md5": "94a6a89d7e2e48d88ea357ce176bccdd",
"sha256": "9c458bdf8e719ed0898052d4b29c47c5530861b4f864a0dcd20d36dfa00dd14f"
},
"downloads": -1,
"filename": "reasonchain-0.2.16.tar.gz",
"has_sig": false,
"md5_digest": "94a6a89d7e2e48d88ea357ce176bccdd",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.7",
"size": 7008138,
"upload_time": "2025-01-04T20:34:20",
"upload_time_iso_8601": "2025-01-04T20:34:20.340566Z",
"url": "https://files.pythonhosted.org/packages/08/f4/a77dd2d1c896147968be8096c70d3697cc36d8ec6d63c6542daa00845f37/reasonchain-0.2.16.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2025-01-04 20:34:20",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "sunnybedi990",
"github_project": "reasonchain",
"travis_ci": false,
"coveralls": false,
"github_actions": false,
"requirements": [
{
"name": "numpy",
"specs": [
[
"<",
"2.0.0"
],
[
">=",
"1.21.0"
]
]
},
{
"name": "scipy",
"specs": [
[
">=",
"1.8.0"
],
[
"<",
"1.14.0"
]
]
},
{
"name": "scikit-learn",
"specs": [
[
">=",
"1.0.0"
]
]
},
{
"name": "torch",
"specs": []
},
{
"name": "tqdm",
"specs": []
},
{
"name": "faiss-cpu",
"specs": []
},
{
"name": "pymilvus",
"specs": []
},
{
"name": "pinecone",
"specs": []
},
{
"name": "qdrant-client",
"specs": []
},
{
"name": "weaviate-client",
"specs": []
},
{
"name": "transformers",
"specs": []
},
{
"name": "ollama",
"specs": []
},
{
"name": "groq",
"specs": []
},
{
"name": "openai",
"specs": []
},
{
"name": "matplotlib",
"specs": []
},
{
"name": "tabula-py",
"specs": []
},
{
"name": "camelot-py",
"specs": []
},
{
"name": "pymupdf",
"specs": []
},
{
"name": "sentence-transformers",
"specs": []
},
{
"name": "tensorflow_hub",
"specs": []
},
{
"name": "gensim",
"specs": []
},
{
"name": "layoutparser",
"specs": []
},
{
"name": "pdf2image",
"specs": []
},
{
"name": "pytesseract",
"specs": []
},
{
"name": "pdfplumber",
"specs": []
},
{
"name": "fastapi",
"specs": []
},
{
"name": "jpype1",
"specs": []
},
{
"name": "llama-index-core",
"specs": [
[
"==",
"0.12.2"
]
]
},
{
"name": "llama-parse",
"specs": []
},
{
"name": "llama-index-readers-file",
"specs": []
},
{
"name": "python-dotenv",
"specs": []
},
{
"name": "opencv-python",
"specs": []
},
{
"name": "datasets",
"specs": []
},
{
"name": "python-pptx",
"specs": []
},
{
"name": "moviepy",
"specs": []
},
{
"name": "SpeechRecognition",
"specs": []
},
{
"name": "ebooklib",
"specs": []
},
{
"name": "bs4",
"specs": []
},
{
"name": "python-docx",
"specs": []
},
{
"name": "whisper",
"specs": []
}
],
"lcname": "reasonchain"
}
JSON
