raqa 1.0.2__tar.gz → 3.0.0__tar.gz

{raqa-1.0.2/src/raqa.egg-info → raqa-3.0.0}/PKG-INFO

@@ -1,6 +1,7 @@
 Metadata-Version: 2.4
 Name: raqa
-Version: 1.0.2
+Version: 3.0.0
+Summary: RAG-powered document Q&A — Streamlit + OpenAI Agents SDK
 Author-email: Jordi Carrera Ventura <jordi.carrera.ventura@gmail.com>
 Project-URL: GitHub repository, https://github.com/JordiCarreraVentura/raqa
 Project-URL: PyPI, https://pypi.org/project/raqa
@@ -10,14 +11,11 @@ Requires-Python: >=3.10
 Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: build
-Requires-Dist: faiss-cpu
-Requires-Dist: joblib
 Requires-Dist: numpy
 Requires-Dist: openai
-Requires-Dist: python-frontmatter
-Requires-Dist: sentence-transformers
-Requires-Dist: tqdm
-Requires-Dist: typer[all]
+Requires-Dist: openai-agents
+Requires-Dist: python-dotenv
+Requires-Dist: streamlit
 Requires-Dist: twine
 Dynamic: license-file
 
@@ -53,7 +51,7 @@ pip install raqa
 
 3. One-shot retrieval
 
-    `python cli.py search "what is retrieval augmented generation?"`
+    `python cli.py search DATABASE_NAME "what is retrieval augmented generation?"`
 
 4. Rebuild and chat
 
@@ -73,8 +71,8 @@ pip install raqa
 ```
 raqa build DATABASE_NAME PATH/TO/FOLDER/WITH/MARKDOWNS
 raqa chat DATABASE_NAME
-raqa search "what is RAG?"
-raqa list
+raqa search DATABASE_NAME "what is RAG?"
+raqa list (DATABASE_NAME)
 raqa stats (DATABASE_NAME)
 raqa rebuild-and-chat DATABASE_NAME PATH/TO/FOLDER/WITH/MARKDOWNS
 ```

{raqa-1.0.2 → raqa-3.0.0}/README.md

@@ -30,7 +30,7 @@ pip install raqa
 
 3. One-shot retrieval
 
-    `python cli.py search "what is retrieval augmented generation?"`
+    `python cli.py search DATABASE_NAME "what is retrieval augmented generation?"`
 
 4. Rebuild and chat
 
@@ -50,8 +50,8 @@ pip install raqa
 ```
 raqa build DATABASE_NAME PATH/TO/FOLDER/WITH/MARKDOWNS
 raqa chat DATABASE_NAME
-raqa search "what is RAG?"
-raqa list
+raqa search DATABASE_NAME "what is RAG?"
+raqa list (DATABASE_NAME)
 raqa stats (DATABASE_NAME)
 raqa rebuild-and-chat DATABASE_NAME PATH/TO/FOLDER/WITH/MARKDOWNS
 ```

{raqa-1.0.2 → raqa-3.0.0}/pyproject.toml

@@ -4,11 +4,11 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "raqa"
-version = "1.0.2"
+version = "3.0.0"
 authors = [
   { name="Jordi Carrera Ventura", email="jordi.carrera.ventura@gmail.com" },
 ]
-description = ""
+description = "RAG-powered document Q&A — Streamlit + OpenAI Agents SDK"
 readme = "README.md"
 requires-python = ">=3.10"
 classifiers = [
@@ -17,14 +17,11 @@ classifiers = [
 ]
 dependencies = [
   "build",
-  "faiss-cpu",
-  "joblib",
   "numpy",
   "openai",
-  "python-frontmatter",
-  "sentence-transformers",
-  "tqdm",
-  "typer[all]",
+  "openai-agents",
+  "python-dotenv",
+  "streamlit",
   "twine"
 ]
 
@@ -33,4 +30,4 @@ dependencies = [
 "PyPI" = "https://pypi.org/project/raqa"
 
 [project.scripts]
-raqa = "raqa.cli:app"
+raqa = "raqa.cli:main"

raqa-3.0.0/src/raqa/__main__.py

@@ -0,0 +1,3 @@
+from .cli import main
+
+main()

raqa-3.0.0/src/raqa/_agent.py

@@ -0,0 +1,63 @@
+import numpy as np
+from dotenv import load_dotenv
+from openai import OpenAI
+from agents import Agent, function_tool
+
+from ._data import load_documents
+
+
+load_dotenv()
+
+_client = None
+chunks = []
+embeddings = None
+
+
+def _get_client() -> OpenAI:
+    global _client
+    if _client is None:
+        _client = OpenAI()
+    return _client
+
+
+def initialize(data_dir: str = "data"):
+    global chunks, embeddings
+    chunks = load_documents(data_dir)
+    if not chunks:
+        embeddings = np.array([], dtype="float32").reshape(0, 0)
+        return
+
+    texts = [c["text"] for c in chunks]
+    response = _get_client().embeddings.create(input=texts, model="text-embedding-3-small")
+    embeddings = np.array([d.embedding for d in response.data], dtype="float32")
+
+
+@function_tool
+def search_docs(query: str) -> str:
+    """Search the loaded documents for information relevant to the query."""
+    if not chunks or embeddings.size == 0:
+        return "No documents loaded."
+
+    q_emb = _get_client().embeddings.create(
+        input=[query], model="text-embedding-3-small"
+    ).data[0].embedding
+    q_vec = np.array(q_emb, dtype="float32")
+
+    scores = embeddings @ q_vec
+    top_k = min(5, len(chunks))
+    indices = np.argsort(scores)[-top_k:][::-1]
+
+    results = [f"[{chunks[i]['source']}]\n{chunks[i]['text']}" for i in indices]
+    return "\n\n---\n\n".join(results)
+
+
+def create_agent() -> Agent:
+    return Agent(
+        name="RAQA",
+        instructions=(
+            "You are a helpful assistant. "
+            "Answer questions based on the provided documents. "
+            "If the search results are not relevant, say so."
+        ),
+        tools=[search_docs],
+    )

raqa-3.0.0/src/raqa/_app.py

@@ -0,0 +1,41 @@
+import os
+import asyncio
+
+import streamlit as st
+from agents import Runner
+
+from raqa._agent import initialize, create_agent
+
+
+st.set_page_config(page_title="RAQA", page_icon="📚", layout="centered")
+
+data_dir = os.environ.get("RAQA_DATA_DIR", "data")
+
+if "ready" not in st.session_state:
+    with st.spinner("Loading documents..."):
+        initialize(data_dir=data_dir)
+        st.session_state.agent = create_agent()
+        st.session_state.messages = []
+        st.session_state.history = None
+        st.session_state.ready = True
+
+st.title("📚 RAQA")
+
+for msg in st.session_state.messages:
+    with st.chat_message(msg["role"]):
+        st.markdown(msg["content"])
+
+if prompt := st.chat_input("Ask a question about your documents"):
+    st.session_state.messages.append({"role": "user", "content": prompt})
+    with st.chat_message("user"):
+        st.markdown(prompt)
+
+    with st.chat_message("assistant"):
+        with st.spinner("Thinking..."):
+            history = st.session_state.history
+            input_for = history + [{"role": "user", "content": prompt}] if history else prompt
+            result = asyncio.run(Runner.run(st.session_state.agent, input_for))
+            st.session_state.history = result.to_input_list()
+            response = result.final_output
+        st.markdown(response)
+    st.session_state.messages.append({"role": "assistant", "content": response})

raqa-3.0.0/src/raqa/_data.py

@@ -0,0 +1,48 @@
+import csv
+import re
+from pathlib import Path
+from typing import List, Dict
+
+
+def load_documents(data_dir: str = "data") -> List[Dict]:
+    chunks = []
+    data_path = Path(data_dir)
+    if not data_path.exists():
+        print(f"⚠️  Data directory '{data_dir}' not found.")
+        return chunks
+
+    for path in sorted(data_path.rglob("*")):
+        if path.suffix == ".md":
+            text = path.read_text(encoding="utf-8")
+            clean = _remove_markdown(text)
+            paras = [p.strip() for p in re.split(r'\n\s*\n', clean) if p.strip()]
+            for i, para in enumerate(paras):
+                chunks.append({"text": para, "source": str(path), "index": i})
+        elif path.suffix == ".txt":
+            text = path.read_text(encoding="utf-8")
+            paras = [p.strip() for p in re.split(r'\n\s*\n', text) if p.strip()]
+            for i, para in enumerate(paras):
+                chunks.append({"text": para, "source": str(path), "index": i})
+        elif path.suffix == ".csv":
+            with open(path, newline="", encoding="utf-8") as f:
+                rows = list(csv.reader(f))
+            if rows:
+                lines = [" | ".join(r) for r in rows]
+                chunks.append({"text": "\n".join(lines), "source": str(path), "index": 0})
+
+    return chunks
+
+
+def _remove_markdown(text: str) -> str:
+    text = re.sub(r'!\[[^\]]*\]\([^\)]+\)', '', text)
+    text = re.sub(r'\[([^\]]+)\]\([^\)]+\)', r'\1', text)
+    text = re.sub(r'(\*\*|__)(.*?)\1', r'\2', text)
+    text = re.sub(r'(\*|_)(.*?)\1', r'\2', text)
+    text = re.sub(r'`([^`]+)`', r'\1', text)
+    text = re.sub(r'^\s*#+\s+(.*)$', r'\1', text, flags=re.MULTILINE)
+    text = re.sub(r'^\s*>\s+', '', text, flags=re.MULTILINE)
+    text = re.sub(r'^\s*[-*_]{3,}\s*$', '', text, flags=re.MULTILINE)
+    text = re.sub(r'^\s*[\-\*]\s+', '', text, flags=re.MULTILINE)
+    text = re.sub(r'^\s*\d+\.\s+', '', text, flags=re.MULTILINE)
+    text = re.sub(r'\n{3,}', '\n\n', text)
+    return text.strip()

raqa-3.0.0/src/raqa/cli.py

@@ -0,0 +1,10 @@
+import sys
+from pathlib import Path
+
+
+def main():
+    """Launch the RAQA Streamlit app."""
+    app_path = Path(__file__).parent / "_app.py"
+    from streamlit.web import cli as stcli
+    sys.argv = ["streamlit", "run", str(app_path)]
+    sys.exit(stcli.main())

{raqa-1.0.2 → raqa-3.0.0/src/raqa.egg-info}/PKG-INFO

@@ -1,6 +1,7 @@
 Metadata-Version: 2.4
 Name: raqa
-Version: 1.0.2
+Version: 3.0.0
+Summary: RAG-powered document Q&A — Streamlit + OpenAI Agents SDK
 Author-email: Jordi Carrera Ventura <jordi.carrera.ventura@gmail.com>
 Project-URL: GitHub repository, https://github.com/JordiCarreraVentura/raqa
 Project-URL: PyPI, https://pypi.org/project/raqa
@@ -10,14 +11,11 @@ Requires-Python: >=3.10
 Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: build
-Requires-Dist: faiss-cpu
-Requires-Dist: joblib
 Requires-Dist: numpy
 Requires-Dist: openai
-Requires-Dist: python-frontmatter
-Requires-Dist: sentence-transformers
-Requires-Dist: tqdm
-Requires-Dist: typer[all]
+Requires-Dist: openai-agents
+Requires-Dist: python-dotenv
+Requires-Dist: streamlit
 Requires-Dist: twine
 Dynamic: license-file
 
@@ -53,7 +51,7 @@ pip install raqa
 
 3. One-shot retrieval
 
-    `python cli.py search "what is retrieval augmented generation?"`
+    `python cli.py search DATABASE_NAME "what is retrieval augmented generation?"`
 
 4. Rebuild and chat
 
@@ -73,8 +71,8 @@ pip install raqa
 ```
 raqa build DATABASE_NAME PATH/TO/FOLDER/WITH/MARKDOWNS
 raqa chat DATABASE_NAME
-raqa search "what is RAG?"
-raqa list
+raqa search DATABASE_NAME "what is RAG?"
+raqa list (DATABASE_NAME)
 raqa stats (DATABASE_NAME)
 raqa rebuild-and-chat DATABASE_NAME PATH/TO/FOLDER/WITH/MARKDOWNS
 ```

{raqa-1.0.2 → raqa-3.0.0}/src/raqa.egg-info/SOURCES.txt

@@ -2,11 +2,11 @@ LICENSE
 README.md
 pyproject.toml
 src/raqa/__init__.py
-src/raqa/agent.py
+src/raqa/__main__.py
+src/raqa/_agent.py
+src/raqa/_app.py
+src/raqa/_data.py
 src/raqa/cli.py
-src/raqa/config.py
-src/raqa/db.py
-src/raqa/utils.py
 src/raqa.egg-info/PKG-INFO
 src/raqa.egg-info/SOURCES.txt
 src/raqa.egg-info/dependency_links.txt

raqa-3.0.0/src/raqa.egg-info/entry_points.txt

@@ -0,0 +1,2 @@
+[console_scripts]
+raqa = raqa.cli:main

raqa-3.0.0/src/raqa.egg-info/requires.txt

@@ -0,0 +1,7 @@
+build
+numpy
+openai
+openai-agents
+python-dotenv
+streamlit
+twine

raqa-1.0.2/src/raqa/agent.py

@@ -1,54 +0,0 @@
-from openai import OpenAI
-
-from .db import VectorDB
-from .utils import get_openai_key
-
-class RAGAgent:
-    def __init__(self, db: VectorDB):
-        self.db = db
-        api_key = get_openai_key()
-        self.client = OpenAI(api_key=api_key)
-
-    def retrieve(self, query: str):
-        results = self.db.search(query)
-        filtered = self.db.nucleus_filter(results)
-
-        context = "\n\n".join(
-            f"[{r['data']['source']}]\n{r['data']['text']}"
-            for r in filtered
-        )
-
-        return context
-
-    def chat(self):
-        print("💬 RAG Agent ready. Type 'exit' to quit.")
-
-        messages = [{"role": "system", "content": "You are a helpful assistant."}]
-
-        while True:
-            user_input = input("\nYou: ")
-            if user_input.lower() in ("exit", "quit"):
-                break
-
-            context = self.retrieve(user_input)
-
-            augmented_prompt = f"""
-Use the context below if relevant:
-
-{context}
-
-User question:
-{user_input}
-"""
-
-            messages.append({"role": "user", "content": augmented_prompt})
-
-            response = self.client.chat.completions.create(
-                model="gpt-4.1-mini",
-                messages=messages
-            )
-
-            reply = response.choices[0].message.content
-            messages.append({"role": "assistant", "content": reply})
-
-            print("\nAssistant:", reply)

raqa-1.0.2/src/raqa/cli.py

@@ -1,148 +0,0 @@
-import os
-from typing import Optional
-
-import typer
-
-from .db import VectorDB
-from .agent import RAGAgent
-from .config import (
-    DB_BASE_DIR,
-    MARKDOWN_ROOT
-)
-
-app = typer.Typer(help="📚 Markdown RAG CLI")
-
-
-# ---------------------------
-# BUILD DATABASE
-# ---------------------------
-@app.command()
-def build(
-    db_name: str = typer.Argument(..., help="Name of the database to create"),
-    markdown_path: str = typer.Argument(MARKDOWN_ROOT, help="Path to markdown files")
-):
-    """Build a database with a user-given name"""
-    db = VectorDB(db_name=db_name)
-    db.build(markdown_path)
-    typer.echo(f"✅ Database '{db_name}' built at {db.db_path}")
-
-# ---------------------------
-# SEARCH ONLY (DEBUG TOOL)
-# ---------------------------
-@app.command()
-def search(
-    db_name: str = typer.Argument(..., help="Database name to search within"),
-    query: str = typer.Argument(..., help="Search query"),
-    k: int = typer.Option(10, help="Top K results")
-):
-    """
-    Run retrieval without LLM (debugging).
-    """
-    db = VectorDB(db_name=db_name)
-    db.load()
-
-    results = db.search(query, k=k)
-
-    typer.secho(f"\n🔎 Raw Results for '{db_name}':\n", fg=typer.colors.BLUE)
-
-    for i, r in enumerate(results):
-        typer.echo(f"\n--- Result {i+1} ---")
-        typer.echo(f"Score: {r['score']:.4f}")
-        typer.echo(f"Source: {r['data']['source']}")
-        typer.echo(r["data"]["text"][:500])
-
-
-# ---------------------------
-# CHAT (MAIN ENTRYPOINT)
-# ---------------------------
-@app.command()
-def chat(
-    db_name: str = typer.Argument("default", help="Database name to use")
-):
-    """Start a chat using a specific database"""
-    db = VectorDB(db_name=db_name)
-    db.load()
-
-    agent = RAGAgent(db=db)
-    agent.chat()
-
-
-# ---------------------------
-# REBUILD + CHAT (CONVENIENCE)
-# ---------------------------
-@app.command()
-def rebuild_and_chat(
-    db_name: str = typer.Argument(..., help="Database name"),
-    markdown_path: str = typer.Argument(..., help="Markdown folder")
-):
-    """
-    Rebuild a named database and immediately start chat.
-    """
-    from .config import DB_EMBEDDINGS_CACHE
-    os.remove(DB_EMBEDDINGS_CACHE)
-    db = VectorDB(db_name=db_name)
-
-    typer.echo(f"🔄 Rebuilding database '{db_name}'...")
-    db.build(markdown_path)
-
-    typer.secho("✅ Build complete. Starting chat...\n", fg=typer.colors.GREEN)
-
-    agent = RAGAgent(db=db)
-    agent.chat()
-
-
-# ---------------------------
-# INSPECT DB
-# ---------------------------
-@app.command()
-def stats(
-    db_name: str = typer.Argument(None, help="Database name (optional)")
-):
-    """
-    Show stats for one or all databases.
-    """
-    if db_name:
-        db = VectorDB(db_name)
-        db.load()
-
-        typer.echo(f"📊 Stats for '{db_name}':")
-        typer.echo(f"Chunks: {len(db.metadata)}")
-        typer.echo(f"Location: {db.db_path}")
-
-    else:
-        typer.echo("📊 All databases:\n")
-
-        for db_path in DB_BASE_DIR.iterdir():
-            if db_path.is_dir():
-                name = db_path.name
-
-                try:
-                    db = VectorDB(name)
-                    db.load()
-                    typer.echo(f"• {name}: {len(db.metadata)} chunks")
-                except Exception:
-                    typer.echo(f"• {name}: ⚠️ corrupted or incomplete")
-
-
-@app.command()
-def list():
-    """
-    List all available databases.
-    """
-    # Fixed relative import for consistency
-    from .config import DB_BASE_DIR
-
-    typer.echo("📚 Available databases:\n")
-
-    found = False
-    for db_path in DB_BASE_DIR.iterdir():
-        if db_path.is_dir():
-            typer.echo(f"• {db_path.name}")
-            found = True
-
-    if not found:
-        typer.echo("No databases found. Use `raqa build` first.")
-        
-
-if __name__ == "__main__":
-    app()

raqa-1.0.2/src/raqa/config.py

@@ -1,27 +0,0 @@
-from pathlib import Path
-
-# User folder for raqa configs
-HOME = Path.home()
-RAQA_HOME = HOME / ".raqa"
-RAQA_HOME.mkdir(parents=True, exist_ok=True)
-
-# OpenAI credential file
-ENV_FILE = RAQA_HOME / "env"
-
-# Default DB base folder (databases will be subfolders)
-DB_BASE_DIR = RAQA_HOME / "databases"
-DB_BASE_DIR.mkdir(parents=True, exist_ok=True)
-DB_EMBEDDINGS_CACHE = RAQA_HOME / "embeddings_cache.json"
-
-# Embeddings & chunk config
-# EMBEDDING_MODEL = "joeddav/xlm-roberta-large-xnli"
-EMBEDDING_MODEL = "text-embedding-3-small"
-CHUNK_WINDOW = 1
-TOP_K = 20
-SIMILARITY_RADIUS = 0.7
-
-# Default markdown folder (can override via CLI)
-MARKDOWN_ROOT = "./markdown_files"
-
-# Ensure directories exist
-RAQA_HOME.mkdir(parents=True, exist_ok=True)

raqa-1.0.2/src/raqa/db.py

@@ -1,197 +0,0 @@
-import json
-import os
-import re
-from pathlib import Path
-from typing import Any
-
-import faiss
-import numpy as np
-import snippyts
-from joblib import Memory
-from dotenv import load_dotenv
-from openai import OpenAI  # Added OpenAI import
-from tqdm import tqdm
-
-from .utils import split_sentences, window_chunks
-from .config import (
-    DB_BASE_DIR,
-    EMBEDDING_MODEL,
-    CHUNK_WINDOW,
-    DB_EMBEDDINGS_CACHE,
-    TOP_K,
-    SIMILARITY_RADIUS
-)
-
-load_dotenv()
-os.environ["TOKENIZERS_PARALLELISM"] = "true"
-
-if os.path.exists(DB_EMBEDDINGS_CACHE):
-    EMBEDDINGS_CACHE = snippyts.from_json(DB_EMBEDDINGS_CACHE)
-else:
-    EMBEDDINGS_CACHE = dict([])
-
-
-def remove_markdown(text: str) -> str:
-    """
-    Removes basic Markdown formatting from a string.
-    """
-    # 1. Remove images: ![alt](url) -> ""
-    text = re.sub(r'!\[[^\]]*\]\([^\)]+\)', '', text)
-    
-    # 2. Remove links but keep the text: [link text](url) -> "link text"
-    text = re.sub(r'\[([^\]]+)\]\([^\)]+\)', r'\1', text)
-    
-    # 3. Remove bold and italics: **text**, __text__, *text*, _text_
-    text = re.sub(r'(\*\*|__)(.*?)\1', r'\2', text)
-    text = re.sub(r'(\*|_)(.*?)\1', r'\2', text)
-    
-    # 4. Remove inline code: `code` -> code
-    text = re.sub(r'`([^`]+)`', r'\1', text)
-    
-    # 5. Remove headers: # Header -> Header
-    text = re.sub(r'^\s*#+\s+(.*)$', r'\1', text, flags=re.MULTILINE)
-    
-    # 6. Remove blockquotes: > quote -> quote
-    text = re.sub(r'^\s*>\s+', '', text, flags=re.MULTILINE)
-    
-    # 7. Remove horizontal rules: ---, ***, ___
-    text = re.sub(r'^\s*[-*_]{3,}\s*$', '', text, flags=re.MULTILINE)
-    
-    # 8. Remove list markers: - item, * item, 1. item
-    text = re.sub(r'^\s*[\-\*]\s+', '', text, flags=re.MULTILINE)
-    text = re.sub(r'^\s*\d+\.\s+', '', text, flags=re.MULTILINE)
-
-    # Clean up excess whitespace/newlines
-    text = re.sub(r'\n{3,}', '\n\n', text)
-    
-    return text.strip()
-
-
-class VectorDB:
-    def __init__(self, db_name: str = "default"):
-        self.name = db_name
-        self.db_path = DB_BASE_DIR / self.name
-        self.db_path.mkdir(parents=True, exist_ok=True)
-
-        # Initialize OpenAI client instead of SentenceTransformer
-        self.client = OpenAI()
-        self.index = None
-        self.metadata = []
-
-
-    def _get_embeddings(self, texts: list[str]) -> np.ndarray:
-        """Fetch embeddings from OpenAI with batching, caching, and robust parsing"""
-        indexed_texts = list(enumerate(texts))
-        
-        # FIX 1: Store the actual vector from cache, not the text string
-        old_texts = [(idx, EMBEDDINGS_CACHE[text]) for idx, text in indexed_texts if text in EMBEDDINGS_CACHE]
-        new_texts = [(idx, text) for idx, text in indexed_texts if text not in EMBEDDINGS_CACHE]
-
-        all_vectors = old_texts
-        batch_size = 100 
-        batches = snippyts.batched(new_texts, batch_size)
-
-        for batch in tqdm(batches, colour="green", desc="Fetching Embeddings"):
-            # Ensure we don't send empty strings to the API
-            clean_batch = [t if t.strip() else " " for _, t in batch]
-            
-            response = self.client.embeddings.create(
-                input=clean_batch,
-                model=EMBEDDING_MODEL
-            )
-
-            res_data = response.data if hasattr(response, 'data') else response
-            
-            for (idx, text), item in zip(batch, res_data):
-                vector = item.embedding if hasattr(item, 'embedding') else item
-                
-                # FIX 2: Handle nested lists returned by some API proxies (e.g. [[val, val...]])
-                if isinstance(vector, list) and len(vector) > 0 and isinstance(vector[0], list):
-                    vector = vector[0]
-                
-                EMBEDDINGS_CACHE[text] = vector
-                all_vectors.append((idx, vector))
-        
-        # Sort back to original order
-        all_vectors.sort(key=lambda x: x[0])
-        final_vectors = [vec for _, vec in all_vectors]
-
-        # FIX 3: Validate dimensions before converting to NumPy
-        if final_vectors:
-            dims = [len(v) for v in final_vectors]
-            if len(set(dims)) > 1:
-                # If this happens, your cache likely has vectors from a different model.
-                # You may need to delete your cache file.
-                raise ValueError(f"Inconsistent embedding dimensions: {set(dims)}. Clear your cache.")
-
-        # Save cache and return matrix
-        snippyts.to_json(EMBEDDINGS_CACHE, DB_EMBEDDINGS_CACHE)
-        return np.array(final_vectors).astype('float32')
-
-
-    def build(self, markdown_root: str):
-        """Ingest markdown files, chunk, encode via OpenAI, and store"""
-        chunks = []
-
-        for path in Path(markdown_root).rglob("*.md"):
-            with open(path, "r", encoding="utf-8") as f:
-                content = f.read()
-                # Use the existing remove_markdown logic
-                clean_text = remove_markdown(content)
-
-            sentences = split_sentences(clean_text)
-            for c in window_chunks(sentences, CHUNK_WINDOW):
-                # Skip chunks that became empty after markdown removal
-                if not c["text"].strip():
-                    continue
-                    
-                chunks.append({
-                    "text": c["text"],
-                    "source": str(path),
-                    "position": c["index"]
-                })
-
-        if not chunks:
-            print("⚠️ No valid text found in markdown files.")
-            return
-
-        texts = [c["text"] for c in chunks]
-        embeddings = self._get_embeddings(texts)
-        
-        # Build FAISS index
-        self.index = faiss.IndexFlatIP(embeddings.shape[1])
-        faiss.normalize_L2(embeddings)
-        self.index.add(embeddings)
-
-        self.metadata = chunks
-        self.save()
-
-
-    def save(self):
-        faiss.write_index(self.index, str(self.db_path / "index.faiss"))
-        with open(self.db_path / "meta.json", "w", encoding="utf-8") as f:
-            json.dump(self.metadata, f)
-
-    def load(self):
-        if not (self.db_path / "index.faiss").exists():
-            raise FileNotFoundError(f"No database found at {self.db_path}. Please build it first.")
-        self.index = faiss.read_index(str(self.db_path / "index.faiss"))
-        with open(self.db_path / "meta.json", "r", encoding="utf-8") as f:
-            self.metadata = json.load(f)
-
-    def search(self, query: str, k=TOP_K):
-        # Encode query using OpenAI
-        q_emb = self._get_embeddings([query])
-        faiss.normalize_L2(q_emb)
-        scores, indices = self.index.search(q_emb, k)
-        return [{"score": float(s), "data": self.metadata[i]} for s, i in zip(scores[0], indices[0])]
-
-    def nucleus_filter(self, results):
-        if not results:
-            return []
-        best = results[0]["score"]
-        filtered = [r for r in results if (best - r["score"]) <= SIMILARITY_RADIUS]
-        scores = np.array([r["score"] for r in filtered])
-        probs = np.exp(scores) / np.sum(np.exp(scores))
-        sampled_indices = np.random.choice(len(filtered), size=min(len(filtered), 10), replace=False, p=probs)
-        return [filtered[i] for i in sampled_indices]

raqa-1.0.2/src/raqa/utils.py

@@ -1,49 +0,0 @@
-import os
-import re
-from pathlib import Path
-from typing import List
-
-from .config import ENV_FILE
-
-def split_sentences(text: str) -> List[str]:
-    # simple but effective
-    sentences = re.split(r'(?<=[.!?])\s+', text)
-    return [s.strip() for s in sentences if s.strip()]
-
-
-def window_chunks(sentences: List[str], window: int = 3):
-    chunks = []
-
-    for i in range(len(sentences)):
-        left = sentences[max(0, i - window): i]
-        center = [sentences[i]]
-        right = sentences[i + 1: i + 1 + window]
-
-        chunk_text = " ".join(left + center + right)
-
-        chunks.append({
-            "text": chunk_text,
-            "center": sentences[i],
-            "index": i
-        })
-
-    return chunks
-
-
-def get_openai_key() -> str:
-    """
-    Load the OpenAI API key from ENV_FILE, prompt user if missing.
-    """
-    if ENV_FILE.exists():
-        key = ENV_FILE.read_text().strip()
-        if key:
-            return key
-
-    # Prompt user
-    print(f"🔑 OpenAI API key not found. Enter your key (it will be saved at {ENV_FILE}):")
-    key = input("API Key: ").strip()
-
-    # Save to file
-    ENV_FILE.write_text(key)
-    print(f"✅ Key saved at {ENV_FILE}")
-    return key

raqa-1.0.2/src/raqa.egg-info/entry_points.txt

@@ -1,2 +0,0 @@
-[console_scripts]
-raqa = raqa.cli:app

raqa-1.0.2/src/raqa.egg-info/requires.txt

@@ -1,10 +0,0 @@
-build
-faiss-cpu
-joblib
-numpy
-openai
-python-frontmatter
-sentence-transformers
-tqdm
-typer[all]
-twine