ragit 0.6__py3-none-any.whl → 0.7.1__py3-none-any.whl

ragit/utils/__init__.py

@@ -0,0 +1,105 @@
+#
+# Copyright RODMENA LIMITED 2025
+# SPDX-License-Identifier: Apache-2.0
+#
+"""
+Ragit utilities module.
+"""
+
+from collections import deque
+from collections.abc import Hashable, Sequence
+from datetime import datetime
+from math import floor
+from typing import Any
+
+import pandas as pd
+
+
+def get_hashable_repr(dct: dict[str, object]) -> tuple[tuple[str, object, float, int | None], ...]:
+    """
+    Returns a hashable representation of the provided dictionary.
+    """
+    queue: deque[tuple[str, object, float, int | None]] = deque((k, v, 0.0, None) for k, v in dct.items())
+    dict_unpacked: list[tuple[str, object, float, int | None]] = []
+    while queue:
+        key, val, lvl, p_ref = queue.pop()
+        if hasattr(val, "items"):  # we have a nested dict
+            dict_unpacked.append((key, "+", lvl, p_ref))
+            if hash(key) != p_ref:
+                lvl += 1
+            queue.extendleft((k, v, lvl, hash(key)) for k, v in val.items())
+        elif isinstance(val, Hashable):
+            dict_unpacked.append((key, val, lvl, p_ref))
+        elif isinstance(val, Sequence):
+            dict_unpacked.append((key, "+", lvl, p_ref))
+            queue.extendleft((key, vv, floor(lvl) + ind * 0.01, hash(key)) for ind, vv in enumerate(val, 1))
+        else:
+            raise ValueError(f"Unsupported type in dict: {type(val)}")
+
+    return tuple(sorted(dict_unpacked, key=lambda it: (it[2], it[0])))
+
+
+def remove_duplicates(items: list[dict[str, Any]]) -> list[dict[str, Any]]:
+    """
+    Deduplicates list of provided dictionary items.
+
+    Parameters
+    ----------
+    items : list[dict]
+        List of items to deduplicate.
+
+    Returns
+    -------
+    list[dict]
+        A deduplicated list of input items.
+    """
+    duplicate_tracker = set()
+    deduplicated_items = []
+    for ind, elem in enumerate(map(get_hashable_repr, items)):
+        if elem not in duplicate_tracker:
+            duplicate_tracker.add(elem)
+            deduplicated_items.append(items[ind])
+    return deduplicated_items
+
+
+def handle_missing_values_in_combinations(df: pd.DataFrame) -> pd.DataFrame:
+    """
+    Handle missing values in experiment data combinations.
+
+    Parameters
+    ----------
+    df : pd.DataFrame
+        Experiment data with combinations being explored.
+
+    Returns
+    -------
+    pd.DataFrame
+        Data with NaN values properly replaced.
+    """
+    if "chunk_overlap" in df.columns:
+        df["chunk_overlap"] = df["chunk_overlap"].map(lambda el: 0 if pd.isna(el) else el)
+
+    return df
+
+
+def datetime_str_to_epoch_time(timestamp: str | int) -> str | int:
+    """
+    Convert datetime string to epoch time.
+
+    Parameters
+    ----------
+    timestamp : str | int
+        Either a datetime string or a unix timestamp.
+
+    Returns
+    -------
+    int
+        Unix timestamp or -1 if parsing fails.
+    """
+    if not isinstance(timestamp, str):
+        return timestamp
+    try:
+        iso_parseable = datetime.fromisoformat(timestamp)
+    except ValueError:
+        return -1
+    return int(iso_parseable.timestamp())

ragit/version.py

@@ -0,0 +1,5 @@
+#
+# Copyright RODMENA LIMITED 2025
+# SPDX-License-Identifier: Apache-2.0
+#
+__version__ = "0.7.1"

ragit-0.7.1.dist-info/METADATA

@@ -0,0 +1,476 @@
+Metadata-Version: 2.4
+Name: ragit
+Version: 0.7.1
+Summary: Automatic RAG Pattern Optimization Engine
+Author: RODMENA LIMITED
+Maintainer-email: RODMENA LIMITED <info@rodmena.co.uk>
+License-Expression: Apache-2.0
+Project-URL: Homepage, https://github.com/rodmena-limited/ragit
+Project-URL: Repository, https://github.com/rodmena-limited/ragit
+Project-URL: Issues, https://github.com/rodmena-limited/ragit/issues
+Keywords: AI,RAG,LLM,GenAI,Optimization,Ollama
+Classifier: Development Status :: 2 - Pre-Alpha
+Classifier: Natural Language :: English
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Programming Language :: Python :: 3.14
+Classifier: Operating System :: MacOS :: MacOS X
+Classifier: Operating System :: POSIX :: Linux
+Requires-Python: <3.14,>=3.12
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: requests>=2.31.0
+Requires-Dist: numpy>=1.26.0
+Requires-Dist: pandas>=2.2.0
+Requires-Dist: pydantic>=2.0.0
+Requires-Dist: python-dotenv>=1.0.0
+Requires-Dist: scikit-learn>=1.5.0
+Requires-Dist: tqdm>=4.66.0
+Provides-Extra: dev
+Requires-Dist: ragit[test]; extra == "dev"
+Requires-Dist: pytest; extra == "dev"
+Requires-Dist: pytest-cov; extra == "dev"
+Requires-Dist: issuedb[web]; extra == "dev"
+Requires-Dist: ruff; extra == "dev"
+Requires-Dist: mypy; extra == "dev"
+Provides-Extra: test
+Requires-Dist: pytest; extra == "test"
+Requires-Dist: pytest-cov; extra == "test"
+Requires-Dist: pytest-mock; extra == "test"
+Dynamic: license-file
+
+# ragit
+
+A Python toolkit for building Retrieval-Augmented Generation (RAG) applications. Ragit provides document loading, chunking, vector search, and LLM integration out of the box, allowing you to build document Q&A systems and code generators with minimal boilerplate.
+
+## Table of Contents
+
+1. [Installation](#installation)
+2. [Configuration](#configuration)
+3. [Tutorial: Using Ragit](#tutorial-using-ragit)
+   - [Loading Documents](#loading-documents)
+   - [The RAGAssistant Class](#the-ragassistant-class)
+   - [Asking Questions](#asking-questions)
+   - [Generating Code](#generating-code)
+   - [Custom Retrieval](#custom-retrieval)
+4. [Tutorial: Platform Integration](#tutorial-platform-integration)
+   - [Flask Integration](#flask-integration)
+   - [FastAPI Integration](#fastapi-integration)
+   - [Command-Line Tools](#command-line-tools)
+   - [Batch Processing](#batch-processing)
+5. [Advanced: Hyperparameter Optimization](#advanced-hyperparameter-optimization)
+6. [API Reference](#api-reference)
+7. [License](#license)
+
+## Installation
+
+```bash
+pip install ragit
+```
+
+Ragit requires an Ollama-compatible API for embeddings and LLM inference. You can use:
+- A local Ollama instance (https://ollama.ai)
+- A cloud-hosted Ollama API
+- Any OpenAI-compatible API endpoint
+
+## Configuration
+
+Ragit reads configuration from environment variables. Create a `.env` file in your project root:
+
+```bash
+# LLM API (cloud or local)
+OLLAMA_BASE_URL=https://your-ollama-api.com
+OLLAMA_API_KEY=your-api-key
+
+# Embedding API (can be different from LLM)
+OLLAMA_EMBEDDING_URL=http://localhost:11434
+
+# Default models
+RAGIT_DEFAULT_LLM_MODEL=llama3.1:8b
+RAGIT_DEFAULT_EMBEDDING_MODEL=mxbai-embed-large
+```
+
+A common setup is to use a cloud API for LLM inference (faster, more capable models) while running embeddings locally (lower latency, no API costs for indexing).
+
+## Tutorial: Using Ragit
+
+This section covers the core functionality of ragit: loading documents, creating a RAG assistant, and querying your knowledge base.
+
+### Loading Documents
+
+Ragit provides several functions for loading and chunking documents.
+
+**Loading a single file:**
+
+```python
+from ragit import load_text
+
+doc = load_text("docs/api-reference.md")
+print(doc.id)       # "api-reference"
+print(doc.content)  # Full file contents
+```
+
+**Loading a directory:**
+
+```python
+from ragit import load_directory
+
+# Load all markdown files
+docs = load_directory("docs/", "*.md")
+
+# Load recursively
+docs = load_directory("docs/", "**/*.md", recursive=True)
+
+# Load multiple file types
+txt_docs = load_directory("docs/", "*.txt")
+rst_docs = load_directory("docs/", "*.rst")
+all_docs = txt_docs + rst_docs
+```
+
+**Custom chunking:**
+
+For fine-grained control over how documents are split:
+
+```python
+from ragit import chunk_text, chunk_by_separator, chunk_rst_sections
+
+# Fixed-size chunks with overlap
+chunks = chunk_text(
+    text,
+    chunk_size=512,      # Characters per chunk
+    chunk_overlap=50,    # Overlap between chunks
+    doc_id="my-doc"
+)
+
+# Split by paragraph
+chunks = chunk_by_separator(text, separator="\n\n")
+
+# Split RST documents by section headers
+chunks = chunk_rst_sections(rst_content, doc_id="tutorial")
+```
+
+### The RAGAssistant Class
+
+The `RAGAssistant` class is the main interface for RAG operations. It handles document indexing, retrieval, and generation in a single object.
+
+```python
+from ragit import RAGAssistant
+
+# Create from a directory
+assistant = RAGAssistant("docs/")
+
+# Create from a single file
+assistant = RAGAssistant("docs/tutorial.rst")
+
+# Create from Document objects
+from ragit import Document
+
+docs = [
+    Document(id="intro", content="Introduction to the API..."),
+    Document(id="auth", content="Authentication uses JWT tokens..."),
+    Document(id="endpoints", content="Available endpoints: /users, /items..."),
+]
+assistant = RAGAssistant(docs)
+```
+
+**Configuration options:**
+
+```python
+assistant = RAGAssistant(
+    "docs/",
+    embedding_model="mxbai-embed-large",  # Model for embeddings
+    llm_model="llama3.1:70b",             # Model for generation
+    chunk_size=512,                        # Characters per chunk
+    chunk_overlap=50,                      # Overlap between chunks
+)
+```
+
+### Asking Questions
+
+The `ask()` method retrieves relevant context and generates an answer:
+
+```python
+assistant = RAGAssistant("docs/")
+
+answer = assistant.ask("How do I authenticate API requests?")
+print(answer)
+```
+
+**Customizing the query:**
+
+```python
+answer = assistant.ask(
+    "How do I authenticate API requests?",
+    top_k=5,                    # Number of chunks to retrieve
+    temperature=0.3,            # Lower = more focused answers
+    system_prompt="You are a technical documentation assistant. "
+                  "Answer concisely and include code examples."
+)
+```
+
+### Generating Code
+
+The `generate_code()` method is optimized for producing clean, runnable code:
+
+```python
+assistant = RAGAssistant("framework-docs/")
+
+code = assistant.generate_code(
+    "Create a REST API endpoint for user registration",
+    language="python"
+)
+print(code)
+```
+
+The output is clean code without markdown formatting. The assistant uses your documentation as context to generate framework-specific, idiomatic code.
+
+### Custom Retrieval
+
+For advanced use cases, you can access the retrieval and generation steps separately:
+
+```python
+assistant = RAGAssistant("docs/")
+
+# Step 1: Retrieve relevant chunks
+results = assistant.retrieve("authentication", top_k=5)
+for chunk, score in results:
+    print(f"Score: {score:.3f}")
+    print(f"Content: {chunk.content[:200]}...")
+    print()
+
+# Step 2: Get formatted context string
+context = assistant.get_context("authentication", top_k=3)
+
+# Step 3: Generate with custom prompt
+prompt = f"""Based on this documentation:
+
+{context}
+
+Write a Python function that validates a JWT token."""
+
+response = assistant.generate(
+    prompt,
+    system_prompt="You are an expert Python developer.",
+    temperature=0.2
+)
+```
+
+## Tutorial: Platform Integration
+
+This section shows how to integrate ragit into web applications and other platforms.
+
+### Flask Integration
+
+```python
+from flask import Flask, request, jsonify
+from ragit import RAGAssistant
+
+app = Flask(__name__)
+
+# Initialize once at startup
+assistant = RAGAssistant("docs/")
+
+@app.route("/ask", methods=["POST"])
+def ask():
+    data = request.get_json()
+    question = data.get("question", "")
+
+    if not question:
+        return jsonify({"error": "question is required"}), 400
+
+    answer = assistant.ask(question, top_k=3)
+    return jsonify({"answer": answer})
+
+@app.route("/search", methods=["GET"])
+def search():
+    query = request.args.get("q", "")
+    top_k = int(request.args.get("top_k", 5))
+
+    results = assistant.retrieve(query, top_k=top_k)
+    return jsonify({
+        "results": [
+            {"content": chunk.content, "score": score}
+            for chunk, score in results
+        ]
+    })
+
+if __name__ == "__main__":
+    app.run(debug=True)
+```
+
+### FastAPI Integration
+
+```python
+from fastapi import FastAPI, HTTPException
+from pydantic import BaseModel
+from ragit import RAGAssistant
+
+app = FastAPI()
+
+# Initialize once at startup
+assistant = RAGAssistant("docs/")
+
+class Question(BaseModel):
+    question: str
+    top_k: int = 3
+    temperature: float = 0.7
+
+class Answer(BaseModel):
+    answer: str
+
+@app.post("/ask", response_model=Answer)
+async def ask(q: Question):
+    if not q.question.strip():
+        raise HTTPException(status_code=400, detail="question is required")
+
+    answer = assistant.ask(
+        q.question,
+        top_k=q.top_k,
+        temperature=q.temperature
+    )
+    return Answer(answer=answer)
+
+@app.get("/search")
+async def search(q: str, top_k: int = 5):
+    results = assistant.retrieve(q, top_k=top_k)
+    return {
+        "results": [
+            {"content": chunk.content, "score": score}
+            for chunk, score in results
+        ]
+    }
+```
+
+### Command-Line Tools
+
+Build CLI tools using argparse or click:
+
+```python
+#!/usr/bin/env python3
+import argparse
+from ragit import RAGAssistant
+
+def main():
+    parser = argparse.ArgumentParser(description="Query documentation")
+    parser.add_argument("question", help="Question to ask")
+    parser.add_argument("--docs", default="docs/", help="Documentation path")
+    parser.add_argument("--top-k", type=int, default=3, help="Context chunks")
+    args = parser.parse_args()
+
+    assistant = RAGAssistant(args.docs)
+    answer = assistant.ask(args.question, top_k=args.top_k)
+    print(answer)
+
+if __name__ == "__main__":
+    main()
+```
+
+Usage:
+
+```bash
+python ask.py "How do I configure logging?"
+python ask.py "What are the API rate limits?" --docs api-docs/ --top-k 5
+```
+
+### Batch Processing
+
+Process multiple questions or generate reports:
+
+```python
+from ragit import RAGAssistant
+
+assistant = RAGAssistant("docs/")
+
+questions = [
+    "What authentication methods are supported?",
+    "How do I handle errors?",
+    "What are the rate limits?",
+]
+
+# Process questions
+results = {}
+for question in questions:
+    results[question] = assistant.ask(question)
+
+# Generate a report
+with open("qa-report.md", "w") as f:
+    f.write("# Documentation Q&A Report\n\n")
+    for question, answer in results.items():
+        f.write(f"## {question}\n\n")
+        f.write(f"{answer}\n\n")
+```
+
+## Advanced: Hyperparameter Optimization
+
+Ragit includes tools to find the optimal RAG configuration for your specific documents and use case.
+
+```python
+from ragit import RagitExperiment, Document, BenchmarkQuestion
+
+# Your documents
+documents = [
+    Document(id="auth", content="Authentication uses Bearer tokens..."),
+    Document(id="api", content="The API supports GET, POST, PUT, DELETE..."),
+]
+
+# Benchmark questions with expected answers
+benchmark = [
+    BenchmarkQuestion(
+        question="What authentication method does the API use?",
+        ground_truth="The API uses Bearer token authentication."
+    ),
+    BenchmarkQuestion(
+        question="What HTTP methods are supported?",
+        ground_truth="GET, POST, PUT, and DELETE methods are supported."
+    ),
+]
+
+# Run optimization
+experiment = RagitExperiment(documents, benchmark)
+results = experiment.run(max_configs=20)
+
+# Get the best configuration
+best = results[0]
+print(f"Best config: chunk_size={best.config.chunk_size}, "
+      f"chunk_overlap={best.config.chunk_overlap}, "
+      f"top_k={best.config.top_k}")
+print(f"Score: {best.score:.3f}")
+```
+
+The experiment tests different combinations of chunk sizes, overlaps, and retrieval parameters to find what works best for your content.
+
+## API Reference
+
+### Document Loading
+
+| Function | Description |
+|----------|-------------|
+| `load_text(path)` | Load a single text file as a Document |
+| `load_directory(path, pattern, recursive=False)` | Load files matching a glob pattern |
+| `chunk_text(text, chunk_size, chunk_overlap, doc_id)` | Split text into overlapping chunks |
+| `chunk_document(doc, chunk_size, chunk_overlap)` | Split a Document into chunks |
+| `chunk_by_separator(text, separator, doc_id)` | Split text by a delimiter |
+| `chunk_rst_sections(text, doc_id)` | Split RST by section headers |
+
+### RAGAssistant
+
+| Method | Description |
+|--------|-------------|
+| `retrieve(query, top_k=3)` | Return list of (Chunk, score) tuples |
+| `get_context(query, top_k=3)` | Return formatted context string |
+| `generate(prompt, system_prompt, temperature)` | Generate text without retrieval |
+| `ask(question, system_prompt, top_k, temperature)` | Retrieve context and generate answer |
+| `generate_code(request, language, top_k, temperature)` | Generate clean code |
+
+### Properties
+
+| Property | Description |
+|----------|-------------|
+| `assistant.num_documents` | Number of loaded documents |
+| `assistant.num_chunks` | Number of indexed chunks |
+| `assistant.embedding_model` | Current embedding model |
+| `assistant.llm_model` | Current LLM model |
+
+## License
+
+Apache-2.0 - RODMENA LIMITED

ragit-0.7.1.dist-info/RECORD

@@ -0,0 +1,18 @@
+ragit/__init__.py,sha256=PjQogIWMlydZFWVECqhmxw-X9i7lEXdUTe2XlT6qYUQ,2213
+ragit/assistant.py,sha256=lXjZRUr_WsYLP3XLOktabgfPVyKOZPdREzyL7cSRufk,11251
+ragit/config.py,sha256=uKLchJQHjH8MImZ2OJahDjSzyasFqgrFb9Z4aHxJ7og,1495
+ragit/loaders.py,sha256=keusuPzXPBiLDVj4hKfPCcge-rm-cnzNRk50fGXvTJs,5571
+ragit/version.py,sha256=eQW9_26GRu30Dsudn-wxi1RA5_L4JTlTW38oZqGpCjo,97
+ragit/core/__init__.py,sha256=j53PFfoSMXwSbK1rRHpMbo8mX2i4R1LJ5kvTxBd7-0w,100
+ragit/core/experiment/__init__.py,sha256=4vAPOOYlY5Dcr2gOolyhBSPGIUxZKwEkgQffxS9BodA,452
+ragit/core/experiment/experiment.py,sha256=Qh1NJkY9LbKaidRfiI8GOwBZqopjK-MSVBuD_JEgO-k,16582
+ragit/core/experiment/results.py,sha256=KHpN3YSLJ83_JUfIMccRPS-q7LEt0S9p8ehDRawk_4k,3487
+ragit/providers/__init__.py,sha256=iliJt74Lt3mFUlKGfSFW-D0cMonUygY6sRZ6lLjeU7M,435
+ragit/providers/base.py,sha256=MJ8mVeXuGWhkX2XGTbkWIY3cVoTOPr4h5XBXw8rAX2Q,3434
+ragit/providers/ollama.py,sha256=Wm2P72mONaFrLXCAauhQqHggtZISFadE06H1s-XKv5E,9279
+ragit/utils/__init__.py,sha256=-UsE5oJSnmEnBDswl-ph0A09Iu8yKNbPhd1-_7Lcb8Y,3051
+ragit-0.7.1.dist-info/licenses/LICENSE,sha256=tAkwu8-AdEyGxGoSvJ2gVmQdcicWw3j1ZZueVV74M-E,11357
+ragit-0.7.1.dist-info/METADATA,sha256=tDEnxgm8bewjWKPlXBHbk7ZR28Do_0MIathdQvHxLNk,13488
+ragit-0.7.1.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+ragit-0.7.1.dist-info/top_level.txt,sha256=pkPbG7yrw61wt9_y_xcLE2vq2a55fzockASD0yq0g4s,6
+ragit-0.7.1.dist-info/RECORD,,

{ragit-0.6.dist-info → ragit-0.7.1.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: setuptools (75.8.0)
+Generator: setuptools (80.9.0)
 Root-Is-Purelib: true
 Tag: py3-none-any