tiny-rag-ai 0.1.0__tar.gz

tiny_rag_ai-0.1.0/PKG-INFO

@@ -0,0 +1,12 @@
+Metadata-Version: 2.4
+Name: tiny-rag-ai
+Version: 0.1.0
+Summary: Complete In-Device RAG chatbot library, no API keys needed.With 2 lines of implementation in applications.
+Project-URL: Homepage, https://github.com/Madhan-1000/tiny-ai
+Requires-Python: >=3.9
+Requires-Dist: faiss-cpu
+Requires-Dist: sentence-transformers
+Requires-Dist: llama-cpp-python
+Requires-Dist: huggingface-hub
+Requires-Dist: pymupdf
+Requires-Dist: numpy

tiny_rag_ai-0.1.0/README.md

@@ -0,0 +1,15 @@
+# Tiny AI
+
+A lightweight Python library for embedding a chatbot into your web app with a simple function call.
+
+## What it is
+- Wraps the Qwen2.5-0.5B-Instruct-GGUF model.
+- Runs inference in about 330 MB of memory on-device.
+- Avoids heavy RAG pipelines by accepting documents directly through its parameters.
+
+## Why use it
+- Minimal setup and small footprint.
+- Focus on your app logic instead of infrastructure.
+
+## Status
+- Code coming soon.

tiny_rag_ai-0.1.0/pyproject.toml

@@ -0,0 +1,28 @@
+[build-system]
+requires = ["setuptools>=64"]
+build-backend = "setuptools.build_meta"    # ← change this line
+
+
+
+[project]
+name = "tiny-rag-ai"
+version = "0.1.0"
+description = "Complete In-Device RAG chatbot library, no API keys needed.With 2 lines of implementation in applications."
+requires-python = ">=3.9"
+
+dependencies = [
+    "faiss-cpu",
+    "sentence-transformers",
+    "llama-cpp-python",
+    "huggingface-hub",
+    "pymupdf",
+    "numpy",
+]
+
+[project.urls]
+Homepage = "https://github.com/Madhan-1000/tiny-ai"
+
+
+
+[tool.setuptools.packages.find]
+where = ["src"]

tiny_rag_ai-0.1.0/setup.cfg

@@ -0,0 +1,4 @@
+[egg_info]
+tag_build = 
+tag_date = 0
+

tiny_rag_ai-0.1.0/src/tiny_rag_ai/__init__.py

@@ -0,0 +1,34 @@
+from . import engine,indexer
+import faiss
+import numpy as np
+
+
+def index(folder_path, save_path="./tiny_ai_data"):
+    """
+    This fucntion does many things to initialize the setup.
+    -> First loads all the documents
+    -> Second the seperates them into chunks 
+    -> Third Loads all the AI models
+    -> Then Embeddes the chunks into vectors
+    -> Then it will store them in a faiss index
+    -> After that the path for saving the indexes is done 
+    """
+    
+    documents=indexer.load_documenmts(folder_path=folder_path)
+    chunks=[]
+    for doc in documents:
+        chunks.extend(indexer.chunk_text(text=doc,chunk_size=500,overlap=50 ))
+    engine._load_models()
+    embeddings=indexer.embed_chunks(chunks=chunks, embed_model=engine._embed_model)
+    dimensions=embeddings.shape[1]
+    faiss_index=faiss.IndexFlatL2(dimensions)
+    engine.set_save_path(save_path) 
+    faiss_index.add(embeddings)
+    indexer.save_index(faiss_index,chunks=chunks,save_path=save_path)
+    
+    
+def chat(query, use_case):  
+    """
+    This fucntion does exactly what is in its name it just calls the ai fucntion.
+    """
+    return engine.chat(user_query=query,usecase=use_case)

tiny_rag_ai-0.1.0/src/tiny_rag_ai/engine.py

@@ -0,0 +1,90 @@
+import os
+import faiss
+import pickle
+from pathlib import Path
+
+
+_cache_dir = os.environ.get("TINY_AI_CACHE_DIR", os.path.expanduser("~/.cache/tiny_ai"))
+_embed_model = None
+_llm = None
+_index = None
+_chunks = None
+_save_path = None
+
+def set_save_path(path: str):
+    global _save_path
+    _save_path = path
+    
+
+def _load_models(cache_dir=_cache_dir):
+    global _embed_model, _llm, _index, _chunks
+    if _embed_model is None:
+        from sentence_transformers import SentenceTransformer
+        _embed_model = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2" ,cache_folder=_cache_dir)
+    if _llm is None:
+        
+        from huggingface_hub import hf_hub_download
+        from llama_cpp import Llama
+        
+        model_path = hf_hub_download(
+            repo_id="Qwen/Qwen2.5-0.5B-Instruct-GGUF",
+            filename="qwen2.5-0.5b-instruct-q4_k_m.gguf",
+            cache_dir=_cache_dir
+        )
+        _llm = Llama(model_path=model_path, n_ctx=4096, n_threads=16)
+
+
+def _load_index():
+    global _index, _chunks ,_save_path
+    if _save_path is None:
+        default = Path("./tiny_ai_data/config.txt")
+        if default.exists():
+            _save_path = default.read_text().strip()
+        else:
+            print("Error: run index() first")
+            return
+    if _index is None:
+        _index = faiss.read_index(str(Path(_save_path) / "faiss.index"))
+    
+    if _chunks is None:
+        try:
+            with open(Path(_save_path) / "chunks.pkl", "rb") as f:
+                _chunks = pickle.load(f)
+        except FileNotFoundError:
+            print("Error: index files not found, run index() first")
+        except Exception as e:
+            print(f"An error occurred: {e}")
+
+def search(query: str, k=6):
+    _load_models()
+    _load_index()
+    query_vector = _embed_model.encode([query]).astype("float32").reshape(1, -1)
+    distances, indices = _index.search(query_vector, k)      
+    results = [_chunks[i] for i in indices[0]]
+    return "\n".join(results)
+
+
+def build_prompt(query: str, context, use_case: str):
+    prompt = f"""<|im_start|>system
+You are a helpful chatbot assistant for {use_case} and you must say I don't know or I can't answer it with a reason why if anything outside of this context.
+Use ONLY the context below to answer.
+CONTEXT:
+{context}<|im_end|>
+<|im_start|>user
+{query}<|im_end|>
+<|im_start|>assistant
+"""
+    return prompt
+
+def generate(prompt:str,max_tokens:int=650):
+    _load_models()
+    answer_output=_llm(prompt,max_tokens=max_tokens,echo=False,stop=["<|im_end|>"])
+    return answer_output["choices"][0]["text"].strip()
+
+def chat(user_query:str,usecase:str):
+    context=search(query=user_query)
+    
+    prompt=build_prompt(user_query,context=context,use_case=usecase)
+    
+    answer=generate(prompt=prompt)
+    return answer

tiny_rag_ai-0.1.0/src/tiny_rag_ai/indexer.py

@@ -0,0 +1,141 @@
+import faiss
+import pickle
+import fitz
+from pathlib import Path
+
+def load_documenmts(folder_path: str) -> list:
+    """
+    Returns all the loaded documents in the path.
+
+    Parameters
+    ----------
+    path : str
+        path for the folder of the web application or any application detailed documents either in .txt files or .pdf files.
+
+    Returns
+    -------
+    returns all the text in the documents in form of array of each document in a string. 
+    """
+    docs=[]
+    folder=Path(folder_path)
+    for pdf_file in folder.glob("**/*.pdf"):
+        
+        pdf_content=load_pdf(pdf_file_path=pdf_file)
+        docs.append(pdf_content)
+    
+    for txt_file in folder.glob("**/*.txt"):
+        
+        txt_file_content=load_txt(txt_file_path=txt_file)
+        docs.append(txt_file_content)
+    
+    return docs
+
+
+def load_txt(txt_file_path: str) -> str:
+    """
+    Returns all the content in the txt file.
+
+    Parameters
+    ----------
+    txt_file_path : str
+        path for the txt file.
+    
+    Returns
+    -------
+    returns the text file content
+    """
+    text=""
+    with open(txt_file_path,"r") as file:
+        text=file.read()
+        
+    return text
+
+
+def load_pdf(pdf_file_path: str) -> str:
+    """
+    Returns all the content in the pdf file.
+
+    Parameters
+    ----------
+    pdf_file_path : str
+        path for the pdf file.
+    
+    Returns
+    -------
+    returns the pdf file content
+    """
+    try:
+        
+        doc=fitz.open(pdf_file_path)
+        text=""
+        
+        for page in doc:
+            text+=page.get_text()
+        doc.close()
+        
+    except Exception as e:
+        print(f"Error processing file {pdf_file_path.name}: {e}")        
+    return text
+
+
+
+def chunk_text(text :str , chunk_size:int , overlap:int)-> list :
+
+    """
+    Returns the chunked text with size of 'chunk_size' and with overlap of 'overlap'.
+
+    Parameters
+    ----------
+    text : str
+          the text of all the documents 
+    chunk_size : int
+          the variable for chunking the data 
+    overlap : int
+          the variable for overlapping the chunked data
+    
+    Returns
+    -------
+    returns all the text in the documents in form of array of each document in a string. 
+    """
+    chunks = []
+    start = 0
+    while start < len(text):
+        end = start + chunk_size
+        chunks.append(text[start:end])
+        start += chunk_size - overlap
+    return chunks
+
+def embed_chunks(chunks:list,embed_model):
+    """
+    Returns the embedded chunks of data.
+
+    Parameters
+    ----------
+    chunks : list
+          the text of all the documents 
+    
+    embed_model
+          the model with transformer
+    Returns
+    -------
+    array of float32 values which are embedded using the model. 
+    """
+    embeddings_array=embed_model.encode(chunks).astype("float32")
+    if len(embeddings_array.shape) == 1:
+        embeddings_array = embeddings_array.reshape(1, -1)
+    return embeddings_array
+
+
+
+def save_index(index, chunks: list, save_path: str):
+    """
+    Saves the FAISS index and chunks to disk.
+    """
+    path = Path(save_path)
+    
+    path.mkdir(parents=True, exist_ok=True)
+    faiss.write_index(index, str(path / "faiss.index"))
+    
+    
+    with open(path / "chunks.pkl", "wb") as f:
+        pickle.dump(chunks,f)

tiny_rag_ai-0.1.0/src/tiny_rag_ai.egg-info/PKG-INFO

@@ -0,0 +1,12 @@
+Metadata-Version: 2.4
+Name: tiny-rag-ai
+Version: 0.1.0
+Summary: Complete In-Device RAG chatbot library, no API keys needed.With 2 lines of implementation in applications.
+Project-URL: Homepage, https://github.com/Madhan-1000/tiny-ai
+Requires-Python: >=3.9
+Requires-Dist: faiss-cpu
+Requires-Dist: sentence-transformers
+Requires-Dist: llama-cpp-python
+Requires-Dist: huggingface-hub
+Requires-Dist: pymupdf
+Requires-Dist: numpy

tiny_rag_ai-0.1.0/src/tiny_rag_ai.egg-info/SOURCES.txt

@@ -0,0 +1,12 @@
+README.md
+pyproject.toml
+src/tiny_rag_ai/__init__.py
+src/tiny_rag_ai/engine.py
+src/tiny_rag_ai/indexer.py
+src/tiny_rag_ai.egg-info/PKG-INFO
+src/tiny_rag_ai.egg-info/SOURCES.txt
+src/tiny_rag_ai.egg-info/dependency_links.txt
+src/tiny_rag_ai.egg-info/requires.txt
+src/tiny_rag_ai.egg-info/top_level.txt
+tests/test_engine.py
+tests/test_indexer.py

tiny_rag_ai-0.1.0/src/tiny_rag_ai.egg-info/dependency_links.txt

@@ -0,0 +1 @@
+

tiny_rag_ai-0.1.0/src/tiny_rag_ai.egg-info/requires.txt

@@ -0,0 +1,6 @@
+faiss-cpu
+sentence-transformers
+llama-cpp-python
+huggingface-hub
+pymupdf
+numpy

tiny_rag_ai-0.1.0/src/tiny_rag_ai.egg-info/top_level.txt

@@ -0,0 +1 @@
+tiny_rag_ai

tiny_rag_ai-0.1.0/tests/test_engine.py

@@ -0,0 +1 @@
+#python script to test engine.py seperatly.

tiny_rag_ai-0.1.0/tests/test_indexer.py

@@ -0,0 +1 @@
+#python script to test indexer.py seperatly.