Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,7 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+rag_system/Try[[:space:]]Docs/Constitution[[:space:]]Of[[:space:]]india.pdf filter=lfs diff=lfs merge=lfs -text
+rag_system/Try[[:space:]]Docs/Contract[[:space:]]of[[:space:]]Insurance.pdf filter=lfs diff=lfs merge=lfs -text
+rag_system/Try[[:space:]]Docs/Indian[[:space:]]Penal[[:space:]]Code.pdf filter=lfs diff=lfs merge=lfs -text
+rag_system/Try[[:space:]]Docs/Origin_of_Species.pdf filter=lfs diff=lfs merge=lfs -text

rag_system/Try Docs/Constitution Of india.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:819ba7adc5e5ae4063f26c82472283ba682ddb662e5a7f864cf26984d99b26ed
+size 2413611

rag_system/Try Docs/Contract of Insurance.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1e9bd70c489bb3e31d73f1df82644c7e0295b736e6177723a13a8e70d434ac3f
+size 452255

rag_system/Try Docs/Indian Penal Code.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:038c736730c09d5b72b1642ab8056607ca546c0b87631811da1a30accd08f81d
+size 1529218

rag_system/Try Docs/Origin_of_Species.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:07baf1c96315e54f042fbb152ac3a4449d70899dc7c2165e03ccf39126a74e20
+size 613465

rag_system/api.py

@@ -21,6 +21,7 @@ from .vector_store import (
     add_documents, load_or_create_store, is_loaded,
     list_collections, get_collection_stats, delete_collection,
     cleanup_stale_collections, get_collection_embedding_mode,
+    pin_collection,
 )
 from .query_engine import query as run_query, stream_query, pipeline_stream_query
 from .eval import evaluate
@@ -46,6 +47,9 @@ _ingest_jobs: dict[str, dict] = {}
 # Raw file bytes for document preview: collection_name -> (bytes, content_type)
 _doc_files: dict[str, tuple[bytes, str]] = {}
 
+# Try Docs file paths: collection_name -> path
+_try_doc_paths: dict[str, "Path"] = {}
+
 _FILE_CONTENT_TYPES: dict[str, str] = {
     '.pdf': 'application/pdf',
     '.txt': 'text/plain; charset=utf-8',
@@ -115,6 +119,41 @@ def _safe_coll_name(filename: str) -> str:
     return safe or 'doc'
 
 
+def _try_doc_collection_name(filename: str) -> str:
+    return f"{settings.try_docs_prefix}{_safe_coll_name(filename)}"
+
+
+def _load_try_docs() -> None:
+    """Load pre-indexed Try Docs into memory and cache raw bytes for preview."""
+    from pathlib import Path
+
+    try_dir = Path(settings.try_docs_path)
+    if not try_dir.exists():
+        logger.info("Try Docs folder not found at %s", try_dir)
+        return
+
+    for path in sorted(try_dir.iterdir()):
+        if not path.is_file():
+            continue
+        suffix = path.suffix.lower()
+        if suffix not in _FILE_CONTENT_TYPES:
+            continue
+
+        collection = _try_doc_collection_name(path.name)
+        _try_doc_paths[collection] = path
+
+        store = load_or_create_store(collection)
+        if store is not None:
+            pin_collection(collection)
+        else:
+            logger.warning("Try Doc index missing for '%s' (%s)", path.name, collection)
+
+        try:
+            _doc_files[collection] = (path.read_bytes(), _FILE_CONTENT_TYPES[suffix])
+        except Exception:
+            logger.warning("Failed to cache Try Doc file bytes for '%s'", path.name)
+
+
 # Lifespan (startup / shutdown)
 @asynccontextmanager
 async def lifespan(app: FastAPI):
@@ -135,6 +174,8 @@ async def lifespan(app: FastAPI):
             if d.is_dir():
                 load_or_create_store(d.name)
 
+    _load_try_docs()
+
     logger.info("RAG API ready!")
 
     # Background session-cleanup loop: remove collections idle > 30 min
@@ -207,6 +248,43 @@ async def embeddings_info():
     return get_embeddings_runtime_info()
 
 
+# ── Try Docs ─────────────────────────────────────────────────────────────────
+
+@app.get("/try_docs", tags=["try_docs"])
+async def list_try_docs():
+    """List pre-indexed Try Docs available to add to a session."""
+    from pathlib import Path
+
+    try_dir = Path(settings.try_docs_path)
+    if not try_dir.exists():
+        return {"docs": []}
+
+    docs = []
+    for path in sorted(try_dir.iterdir()):
+        if not path.is_file():
+            continue
+        suffix = path.suffix.lower()
+        if suffix not in _FILE_CONTENT_TYPES:
+            continue
+
+        collection = _try_doc_collection_name(path.name)
+        _try_doc_paths.setdefault(collection, path)
+
+        index_path = Path(settings.faiss_index_path) / collection
+        stats = get_collection_stats(collection) if index_path.exists() else None
+
+        docs.append({
+            "filename": path.name,
+            "collection": collection,
+            "chunks": stats["chunk_count"] if stats else 0,
+            "embedding_mode": stats["embedding_mode"] if stats else None,
+            "size_mb": stats["size_mb"] if stats else 0.0,
+            "ready": stats is not None,
+        })
+
+    return {"docs": docs}
+
+
 # ── Ingest ────────────────────────────────────────────────────────────────────
 
 @app.post("/ingest", response_model=IngestResponse, tags=["ingest"])
@@ -503,6 +581,14 @@ async def get_document_raw(collection_name: str):
     """Serve raw document bytes for in-browser preview."""
     from fastapi.responses import Response
     entry = _doc_files.get(collection_name)
+    if not entry and collection_name in _try_doc_paths:
+        path = _try_doc_paths[collection_name]
+        try:
+            suffix = path.suffix.lower()
+            _doc_files[collection_name] = (path.read_bytes(), _FILE_CONTENT_TYPES.get(suffix, 'application/octet-stream'))
+            entry = _doc_files.get(collection_name)
+        except Exception:
+            entry = None
     if not entry:
         raise HTTPException(status_code=404, detail=f"Document '{collection_name}' not available for preview")
     data, media_type = entry

rag_system/config.py

@@ -20,6 +20,10 @@ class Settings(BaseSettings):
     embedding_batch_size: int = 32
     embedding_normalize: bool = True
 
+    # Try Docs (pre-indexed demo docs)
+    try_docs_path: str = os.path.join(os.path.dirname(__file__), "Try Docs")
+    try_docs_prefix: str = "try__"
+
     #FAISS
     faiss_index_path: str = "./faiss_indexes"
     faiss_index_name: str = "prod_rag"

rag_system/vector_store.py

@@ -26,6 +26,7 @@ _stores: dict[str, FAISS] = {}
 # Tracks last-used timestamp per collection (epoch seconds) for TTL-based cleanup
 _last_used: dict[str, float] = {}
 _collection_embeddings: dict[str, str] = {}
+_pinned: set[str] = set()
 
 _EMBEDDING_META_FILE = "embedding.json"
 
@@ -64,6 +65,14 @@ def get_collection_embedding_mode(collection: str) -> Optional[str]:
     return None
 
 
+def pin_collection(collection: str) -> None:
+    _pinned.add(collection)
+
+
+def is_pinned_collection(collection: str) -> bool:
+    return collection in _pinned
+
+
 def resolve_embedding_mode_for_collections(
     collections: list[str],
     requested_mode: Optional[str] = None,
@@ -274,7 +283,10 @@ def cleanup_stale_collections(ttl_seconds: int = 1800) -> list[str]:
     Returns the list of collection names that were removed.
     """
     cutoff = time.time() - ttl_seconds
-    stale = [name for name, ts in list(_last_used.items()) if ts < cutoff]
+    stale = [
+        name for name, ts in list(_last_used.items())
+        if ts < cutoff and name not in _pinned
+    ]
     for name in stale:
         logger.info(f"Cleaning up stale collection '{name}' (idle > {ttl_seconds}s)")
         delete_collection(name)
@@ -290,6 +302,8 @@ def delete_collection(collection: str) -> bool:
         del _stores[collection]
     if collection in _collection_embeddings:
         del _collection_embeddings[collection]
+    if collection in _pinned:
+        _pinned.discard(collection)
 
     # Local import to avoid circular dependency with retriever
     from .retriever import _bm25_cache