fix: restore hybrid search and patch MCP/embedding bugs

.env.example

@@ -0,0 +1,3 @@
+HF_TOKEN=
+CONTEXT8_DB_HOST=localhost
+CONTEXT8_DB_PORT=50051

.gitignore

@@ -31,3 +31,6 @@ Thumbs.db
 # Models cache
 .cache/
 models/
+
+
+.env

pyproject.toml

@@ -23,6 +23,7 @@ dependencies = [
     "mcp>=1.0.0",
     "click>=8.0.0",
     "rich>=13.0.0",
+    "ruff>=0.15.11",
 ]
 
 [project.optional-dependencies]

src/context8/cli/commands/ops.py

@@ -7,7 +7,7 @@
 from rich.panel import Panel
 from rich.table import Table
 
-from ...config import COLLECTION_NAME, DB_URL
+from ...config import COLLECTION_NAME, DB_URL, TEXT_EMBED_DIM
 from ..ui import check_actian_sdk, check_db_connection, console
 
 
@@ -142,7 +142,7 @@ def doctor():
                     import actian_vectorai as _av
 
                     _filter = _av.FilterBuilder().must(_av.Field("language").eq("python")).build()
-                    _zero = [0.0] * 384
+                    _zero = [0.0] * TEXT_EMBED_DIM
                     storage.client.points.search(
                         COLLECTION_NAME,
                         vector=_zero,

src/context8/embeddings/service.py

@@ -4,7 +4,7 @@
 import logging
 import os
 
-from ..config import CODE_MODEL, TEXT_MODEL
+from ..config import CODE_EMBED_DIM, CODE_MODEL, TEXT_EMBED_DIM, TEXT_MODEL
 from .tokenizer import BM25Tokenizer
 
 logger = logging.getLogger("context8.embeddings")
@@ -57,7 +57,7 @@ def code_model(self):
         return self._code_model
 
     def _cache_key(self, text: str, model_tag: str) -> str:
-        return hashlib.md5(f"{model_tag}:{text[:500]}".encode()).hexdigest()
+        return hashlib.md5(f"{model_tag}:{text}".encode()).hexdigest()
 
     def _get_cached(self, text: str, model_tag: str) -> list[float] | None:
         return self._cache.get(self._cache_key(text, model_tag))
@@ -68,7 +68,7 @@ def _set_cached(self, text: str, model_tag: str, vector: list[float]) -> None:
 
     def embed_text(self, text: str) -> list[float]:
         if not text.strip():
-            return [0.0] * 384
+            return [0.0] * TEXT_EMBED_DIM
 
         cached = self._get_cached(text, "text")
         if cached is not None:
@@ -81,7 +81,7 @@ def embed_text(self, text: str) -> list[float]:
 
     def embed_code(self, code: str) -> list[float]:
         if not code.strip():
-            dim = 768 if self._use_code_model else 384
+            dim = CODE_EMBED_DIM if self._use_code_model else TEXT_EMBED_DIM
             return [0.0] * dim
 
         cached = self._get_cached(code, "code")

src/context8/mcp/server.py

@@ -1,5 +1,6 @@
 from __future__ import annotations
 
+import asyncio
 import logging
 import threading
 from typing import Any
@@ -25,11 +26,11 @@ async def list_tools() -> list[Tool]:
 async def call_tool(name: str, arguments: dict[str, Any]) -> list[TextContent]:
     try:
         # Try extra tools first (browse, ecosystem)
-        result = call_extra_tool(name, arguments)
+        result = await asyncio.to_thread(call_extra_tool, name, arguments)
         if result is not None:
             return result
         # Fall through to core tools
-        return tools_module.call_tool(name, arguments)
+        return await asyncio.to_thread(tools_module.call_tool, name, arguments)
     except Exception as e:
         logger.error(f"Tool '{name}' failed: {e}", exc_info=True)
         return [TextContent(type="text", text=f"Context8 error: {str(e)}")]
@@ -50,6 +51,4 @@ async def run_server():
 
 
 if __name__ == "__main__":
-    import asyncio
-
     asyncio.run(run_server())

src/context8/mcp/tools_browse.py

@@ -97,39 +97,40 @@ def _handle_browse(args: dict) -> list[TextContent]:
     from ..storage import StorageService
 
     storage = StorageService()
+    try:
+        records = browse(
+            storage,
+            tag=args.get("tag"),
+            language=args.get("language"),
+            framework=args.get("framework"),
+            error_type=args.get("error_type"),
+            limit=args.get("limit", 20),
+        )
+
+        if not records:
+            return [TextContent(type="text", text="No records match those filters.")]
+
+        lines = [f"Found {len(records)} record(s):\n"]
+        for i, r in enumerate(records, 1):
+            meta = []
+            if r.language:
+                meta.append(r.language)
+            if r.framework:
+                meta.append(r.framework)
+            if r.error_type:
+                meta.append(r.error_type)
+            meta_str = f" ({', '.join(meta)})" if meta else ""
+
+            lines.append(f"[{i}] {r.problem_text[:120]}{meta_str}")
+            lines.append(f"    Fix: {r.solution_text[:150]}")
+            if r.tags:
+                lines.append(f"    Tags: {', '.join(r.tags[:5])}")
+            lines.append(f"    ID: {r.id}  Confidence: {r.confidence:.0%}")
+            lines.append("")
 
-    records = browse(
-        storage,
-        tag=args.get("tag"),
-        language=args.get("language"),
-        framework=args.get("framework"),
-        error_type=args.get("error_type"),
-        limit=args.get("limit", 20),
-    )
-
-    if not records:
-        return [TextContent(type="text", text="No records match those filters.")]
-
-    lines = [f"Found {len(records)} record(s):\n"]
-    for i, r in enumerate(records, 1):
-        meta = []
-        if r.language:
-            meta.append(r.language)
-        if r.framework:
-            meta.append(r.framework)
-        if r.error_type:
-            meta.append(r.error_type)
-        meta_str = f" ({', '.join(meta)})" if meta else ""
-
-        lines.append(f"[{i}] {r.problem_text[:120]}{meta_str}")
-        lines.append(f"    Fix: {r.solution_text[:150]}")
-        if r.tags:
-            lines.append(f"    Tags: {', '.join(r.tags[:5])}")
-        lines.append(f"    ID: {r.id}  Confidence: {r.confidence:.0%}")
-        lines.append("")
-
-    storage.close()
-    return [TextContent(type="text", text="\n".join(lines))]
+        return [TextContent(type="text", text="\n".join(lines))]
+    finally:
+        storage.close()
 
 
 def _handle_ecosystem(args: dict) -> list[TextContent]:

src/context8/search/engine.py

@@ -299,11 +299,11 @@ def _search_sparse(
         VectorAIError = av.exceptions.VectorAIError
 
         try:
+            sparse_vec = av.SparseVector(indices=indices, values=values)
             return self.storage.client.points.search(
                 COLLECTION_NAME,
-                vector=values,
-                vector_name="keywords",
-                sparse_indices=indices,
+                vector=sparse_vec,
+                using="keywords",
                 filter=search_filter,
                 limit=limit,
                 with_payload=True,

src/context8/storage.py

@@ -103,8 +103,8 @@ def initialize(self) -> bool:
     def sparse_supported(self) -> bool:
         if self._sparse_supported is None:
             try:
-                self.client.collections.get_info(COLLECTION_NAME)
-                self._sparse_supported = False
+                info = self.client.collections.get_info(COLLECTION_NAME)
+                self._sparse_supported = bool(self._discover_sparse_vectors(info))
             except Exception:
                 self._sparse_supported = False
         return self._sparse_supported