codebase-retrieval-context-engine 2.0.1__py3-none-any.whl → 2.0.3__py3-none-any.whl

codebase_retrieval_context_engine-2.0.3.dist-info/METADATA

@@ -0,0 +1,107 @@
+Metadata-Version: 2.4
+Name: codebase-retrieval-context-engine
+Version: 2.0.3
+Summary: Code retrieval engine — hybrid embedding + graph search for LLM context injection.
+Project-URL: Homepage, https://github.com/nullmastermind/local-context-engine
+Project-URL: Repository, https://github.com/nullmastermind/local-context-engine
+Project-URL: Issues, https://github.com/nullmastermind/local-context-engine/issues
+Author: nullmastermind
+License: Apache-2.0
+License-File: LICENSE
+Keywords: LLM,MCP,code-search,codebase-retrieval,context,embeddings,retrieval
+Classifier: License :: OSI Approved :: Apache Software License
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: Topic :: Software Development :: Libraries
+Requires-Python: >=3.11
+Requires-Dist: mcp>=1.1.2
+Requires-Dist: numpy>=2.0
+Requires-Dist: pathspec>=0.11
+Requires-Dist: pydantic>=2.0
+Requires-Dist: python-dotenv>=1.0
+Requires-Dist: rich>=13.0
+Requires-Dist: typer>=0.12
+Provides-Extra: anthropic
+Requires-Dist: anthropic>=0.25; extra == 'anthropic'
+Provides-Extra: aws
+Requires-Dist: boto3>=1.34; extra == 'aws'
+Provides-Extra: azure
+Requires-Dist: openai>=1.0; extra == 'azure'
+Provides-Extra: debug
+Requires-Dist: gradio>=4.0; extra == 'debug'
+Provides-Extra: dev
+Requires-Dist: httpx; extra == 'dev'
+Requires-Dist: mypy; extra == 'dev'
+Requires-Dist: pytest-asyncio; extra == 'dev'
+Requires-Dist: pytest-cov; extra == 'dev'
+Requires-Dist: pytest>=8.0; extra == 'dev'
+Requires-Dist: respx; extra == 'dev'
+Requires-Dist: ruff; extra == 'dev'
+Provides-Extra: gcp
+Requires-Dist: anthropic[vertex]>=0.25; extra == 'gcp'
+Requires-Dist: google-cloud-aiplatform>=1.38; extra == 'gcp'
+Provides-Extra: google
+Requires-Dist: google-genai>=2.7.0; extra == 'google'
+Provides-Extra: openai
+Requires-Dist: openai>=1.0; extra == 'openai'
+Provides-Extra: treesitter
+Requires-Dist: tree-sitter-c-sharp>=0.21; extra == 'treesitter'
+Requires-Dist: tree-sitter-go>=0.21; extra == 'treesitter'
+Requires-Dist: tree-sitter-java>=0.21; extra == 'treesitter'
+Requires-Dist: tree-sitter-javascript>=0.21; extra == 'treesitter'
+Requires-Dist: tree-sitter-php>=0.21; extra == 'treesitter'
+Requires-Dist: tree-sitter-python>=0.21; extra == 'treesitter'
+Requires-Dist: tree-sitter-ruby>=0.21; extra == 'treesitter'
+Requires-Dist: tree-sitter-rust>=0.21; extra == 'treesitter'
+Requires-Dist: tree-sitter-typescript>=0.21; extra == 'treesitter'
+Requires-Dist: tree-sitter>=0.21; extra == 'treesitter'
+Provides-Extra: voyage
+Requires-Dist: voyageai>=0.3; extra == 'voyage'
+Description-Content-Type: text/markdown
+
+<div align="center">
+  <h1>codebase-retrieval-context-engine</h1>
+  <p><strong>Code retrieval engine for LLM context via MCP.</strong></p>
+  <p>
+    <a href="LICENSE"><img src="https://img.shields.io/badge/License-Apache_2.0-blue.svg" alt="License"/></a>
+  </p>
+</div>
+
+---
+
+## Add to Claude Code
+
+```bash
+claude mcp add codebase-retrieval -e CORBELL_LLM_PROVIDER=google -e GOOGLE_API_KEY=your-google-api-key -e GOOGLE_MODEL=gemini-3.1-flash-lite -e CORBELL_EMBEDDING_MODEL=voyage-4-lite -e VOYAGE_API_KEY=your-voyage-api-key -- uvx codebase-retrieval-context-engine
+```
+
+That's it. The AI agent passes workspace path and triggers index builds automatically.
+
+---
+
+## Build index manually (optional)
+
+```bash
+uvx codebase-retrieval-context-engine index build
+```
+
+Run from your project root. Env vars (`CORBELL_LLM_PROVIDER`, `GOOGLE_API_KEY`, etc.) must be set in your shell.
+
+---
+
+## Environment variables
+
+| Variable | Description |
+|---|---|
+| `CORBELL_LLM_PROVIDER` | LLM provider for reranking (`google`, `anthropic`, `openai`) |
+| `GOOGLE_API_KEY` | Google AI API key (supports multiple: `key1,key2,key3`) |
+| `GOOGLE_MODEL` | e.g. `gemini-3.1-flash-lite` |
+| `CORBELL_EMBEDDING_MODEL` | `voyage-4-lite`, `voyage-code-3`, or `gemini-embedding-001` |
+| `VOYAGE_API_KEY` | Voyage AI API key (supports multiple: `key1,key2,key3`). Add a card to billing to unlock rate limits. |
+
+---
+
+## License
+
+Apache 2.0

{codebase_retrieval_context_engine-2.0.1.dist-info → codebase_retrieval_context_engine-2.0.3.dist-info}/RECORD

@@ -1,46 +1,47 @@
-corbell/__init__.py,sha256=gUE7pWjjfUbFYz86nNIEp6iM5c3J-2s84OydWPFmEMc,124
+corbell/__init__.py,sha256=BaFS4Y0-zDapO7abzb83G5HBc7cB1xLntwgXWJbDixk,124
 corbell/cli/__init__.py,sha256=5-MP6JIWgp4nDLNIhqP6Gtx97GESaIYg3NGxtRGaMv0,28
-corbell/cli/main.py,sha256=anYpXiyQD6_1wMS0Dtef6Rxtxd0NEFe7HHnerHxf3J4,1835
+corbell/cli/main.py,sha256=CP5EHizFLaBLF1EohgVo_-XFlm4VaO6peQaSnzyfxAI,1954
 corbell/cli/commands/__init__.py,sha256=0mAOs3RWC7XMZnGRN677hjPCHHQKDq9ASjIr_GQM3js,37
-corbell/cli/commands/index.py,sha256=pdtGr0dNy3OdxqMBwUjkhOXxxR0N966T3MqIGweZHQI,3021
+corbell/cli/commands/debug.py,sha256=wdwveCeQSgcQbNg5-R5ekU_smEQKMq8WfH0obBbq3i8,10764
+corbell/cli/commands/index.py,sha256=_nv5TC2O1xusX2gY8s2p00xPLN3wQrHEiFmc0EL6oHY,3432
 corbell/cli/commands/query.py,sha256=Sh-xnVj4n3zAI2hTxVyMTqFEPsq3vkWucfljnCEaGyU,2310
 corbell/core/__init__.py,sha256=VS9PnhHr4NXYlWs1TLCyllnVCNsiwVZ1Xj-AOBhZpAU,29
-corbell/core/constants.py,sha256=HTGYpShlp9pP2_a4WngHtTujUQfHcypFAYoaczmkBdQ,1061
-corbell/core/gitignore.py,sha256=VS7_s6NwZWQAwgLiaRzPHdBRIj86XdnPm_P_x_e0hvI,2266
-corbell/core/llm_client.py,sha256=2MDwe6kr_EyY3DFv3fNO91WCig8ER021ogzdLGH3IN8,26219
-corbell/core/workspace.py,sha256=v9H56MxhI2iHPr3P9fU5DnIiniGehozWScfcd6TG2gQ,14189
+corbell/core/constants.py,sha256=P0fCJ0J5V2Nt348ZAVH1bHd9dFPJRLtpUyQhHPAl0_8,1203
+corbell/core/gitignore.py,sha256=UO588tAxSVv7YEGNDjzdcBys_aqMIAhXrDgToRfcnzc,2347
+corbell/core/llm_client.py,sha256=qGKuptxMAMDwqvhGAKVjppf2p-sX-auaA26WKo6Nlkk,26221
+corbell/core/workspace.py,sha256=p24p_yJss7B3UPbv7Qx7XCUagJ2YKTrsBxDhFLCfqd4,14118
 corbell/core/embeddings/__init__.py,sha256=RCekvfNkFuMGEDLnls78i3znR84cTdnj4KJ_PeQrMNg,213
 corbell/core/embeddings/base.py,sha256=udPW4XmcPhCpNQA6n8KqMcu2JXvVNv1JjdRJmFq5ZRA,2175
-corbell/core/embeddings/extractor.py,sha256=hOolMX6JX3sVBf062h2zUQpr9SVt81S0hzhNCeJoV1I,7180
+corbell/core/embeddings/extractor.py,sha256=2_BxRpsUcz-C-3HXjvlARqM3U5dzHRJcPR_hhPdMxSE,7314
 corbell/core/embeddings/factory.py,sha256=Lonjbk8Lsxykz-2ZEgFCWoH9zZ005Qm4dXVdA6P4qJY,1817
-corbell/core/embeddings/model.py,sha256=5guf3qpqOaC7fSp-OZlQjBUQt7jOf_3eE3w6X6yNY2I,15224
+corbell/core/embeddings/model.py,sha256=hU-SyW7YM9jGv9-_-bfxxOUh1ZZdc-8fpDK7o5j5s88,14289
 corbell/core/embeddings/search_cache.py,sha256=FHzO3mu4m4MJGy2jOFwb9GCEypcT11CcVrLts4Ib0ho,3351
-corbell/core/embeddings/sqlite_store.py,sha256=8rv89WOMqMm-JhJO36-FdRiC68Ija3TwHkrmRrPr1os,10158
+corbell/core/embeddings/sqlite_store.py,sha256=99lHU_gPYwKw9BhUMS-XimQI8vDpBbBrIc_RkrsVdOM,11676
 corbell/core/graph/__init__.py,sha256=VaxDKeXMgMEBBMC0dglwj68A_aNYRI5O8VM6oMC1GIM,29
-corbell/core/graph/builder.py,sha256=8BtUVvh_3Sq_XQiYf3yZifRygTrc-_hE770luqwZ3e0,30423
-corbell/core/graph/method_graph.py,sha256=XJpojhBVXon2RNMIaOqE3NgRjUdP0IAnRQuKYGhO5js,49753
+corbell/core/graph/builder.py,sha256=dXUdAhuZ4t-wuW4dFZHz6k9-wBXdYkY6dysjQIkvl3Q,32214
+corbell/core/graph/method_graph.py,sha256=fwmkSZXiGGYZIc2iC-6hbTrb26fAwielOrJBlqaz8Oc,57594
 corbell/core/graph/schema.py,sha256=swy1VZZpL88LPEj6zihl5bglQLrGD-ohOYjFeNC31a0,5253
-corbell/core/graph/sqlite_store.py,sha256=D9que8zHV68x_tiFY-q09Ipk-dXR9LWM3hFPQLGNDto,20019
+corbell/core/graph/sqlite_store.py,sha256=B1ObNit7MXbQpst6dpuloTcFAmUim_MoP3PSCATf_4A,21116
 corbell/core/graph/providers/__init__.py,sha256=__ZVe1uwIHSyFh_t-V4MyT5MsM5hooTOrxxkm9Txt7o,268
 corbell/core/graph/providers/aws_patterns.py,sha256=w2iF5qQJcV7S6J64ZYb3IzGPdXjCc37YX5sNnHz8mXY,2818
 corbell/core/graph/providers/azure_patterns.py,sha256=tJ9AQQXW2xYzJ36wNOxTHHhaivaCv3RYEMJUjw8WjeQ,3515
 corbell/core/graph/providers/gcp_patterns.py,sha256=vIofjanvRWGhFftuGdzt9YgTIGZRJz7lLG0abUNjFdA,2789
 corbell/core/indexing/__init__.py,sha256=VczeSHUfKR3YVowGCleFjo2pIpDHfl9kl-OkEl8szow,47
-corbell/core/indexing/builder.py,sha256=l1Hj9Lrze2zuBvwtauDgmjAWWVvKbKXWKMgtv8uA-TQ,21967
+corbell/core/indexing/builder.py,sha256=apF-FFz_bZ6SeBEVVZzNXMavp9zuLVMVhg4598YJfMs,33333
 corbell/core/indexing/lock.py,sha256=uUMelIrtrp6Ww9rTfbl2OvomByc-IJyiHIMnptfA4xI,4743
-corbell/core/indexing/tracker.py,sha256=mbL1M-EeYf6KoIT5qoz7LCHwSHL6UlZNX7mjm4DczR0,8469
+corbell/core/indexing/tracker.py,sha256=UCeKARiUMyZcg1yvbIZxibZUM2HOA-_6rNTkyPgpQhE,8571
 corbell/core/mcp/__init__.py,sha256=DDzfuVbX_GBTM5Nqy34JVgDUMeFd2_5ZcVMVuvjOddU,32
-corbell/core/mcp/server.py,sha256=nTiPQ9yyenL7uhgLCsGwEm7yyoqk1tUPTsZYFAAmPBU,7270
+corbell/core/mcp/server.py,sha256=CmkqS2EYx4eRzquaJNdPPAx_G07_sJUaK1v_u_aXhTc,5380
 corbell/core/query/__init__.py,sha256=OCyVRZOyh_eLGhOxR_JYyH6zp8O7qy_-rC3fqGHm7Bc,56
-corbell/core/query/diagnostics.py,sha256=ObQyZWmMVRXEHFYGXBP2-EMBmM8SYr0H6cCi95uFnIk,1406
-corbell/core/query/engine.py,sha256=4PRW3R0Wws7BPAqDD4hrWjpqOq8mx9dCrtABWA-Wmus,11635
+corbell/core/query/diagnostics.py,sha256=o9uIAYFQy8hHua1xLMToSaQPP6xcmnvDJMY3fVg1Dhg,2102
+corbell/core/query/engine.py,sha256=wqaZy-ACZQhLua9mlgad4boowDsRFW1TQtQpP5dTReU,17374
 corbell/core/query/enhancer.py,sha256=w5mvm1B8qQZpL6RVhMuhq_rls77hakGSNUyanfkyNEU,3934
 corbell/core/query/formatter.py,sha256=xMr8HE-oxBSEKb514aixY7aoUWGeYoK1w5wnaIlCYEc,2813
 corbell/core/query/graph_expander.py,sha256=Y-yKnr6db-OM2Gh8ukYgVIcUZa6-wfWA-GhdvOwf_yA,9184
 corbell/core/query/merger.py,sha256=fs6PL7X7EweXnSnDRnpzmpaU8JjwJpL0akzm4hSwLJk,6168
-corbell/core/query/reranker.py,sha256=HYckYiUVZ80mbLGHhK4IHxNI7uUqNaztwXLbYgdnoWU,4298
-codebase_retrieval_context_engine-2.0.1.dist-info/METADATA,sha256=lXO0MmvDLFejoeUsHcgOcIRYr284dwjN7YpBpkUloGU,17438
-codebase_retrieval_context_engine-2.0.1.dist-info/WHEEL,sha256=mffPy8wBnZQn2VnJUU5jE99KsxaSfiyMHV9Yt0aLVxs,87
-codebase_retrieval_context_engine-2.0.1.dist-info/entry_points.txt,sha256=vFB4a4Qb7Ty182usK8deJXiis0UYnGIUDusw0V3Jya8,115
-codebase_retrieval_context_engine-2.0.1.dist-info/licenses/LICENSE,sha256=HrhfyXIkWY2tGFK11kg7vPCqhgh5DcxleloqdhrpyMY,11558
-codebase_retrieval_context_engine-2.0.1.dist-info/RECORD,,
+corbell/core/query/reranker.py,sha256=0M8Km2WEO3NX46gT0mF7ma9e0v_HOYXu-t6WgF5U2tI,7262
+codebase_retrieval_context_engine-2.0.3.dist-info/METADATA,sha256=yb84Ich965QFp98h1XcO_uk9uhRn-OaYGG8zsZtEWp0,4089
+codebase_retrieval_context_engine-2.0.3.dist-info/WHEEL,sha256=mffPy8wBnZQn2VnJUU5jE99KsxaSfiyMHV9Yt0aLVxs,87
+codebase_retrieval_context_engine-2.0.3.dist-info/entry_points.txt,sha256=vFB4a4Qb7Ty182usK8deJXiis0UYnGIUDusw0V3Jya8,115
+codebase_retrieval_context_engine-2.0.3.dist-info/licenses/LICENSE,sha256=HrhfyXIkWY2tGFK11kg7vPCqhgh5DcxleloqdhrpyMY,11558
+codebase_retrieval_context_engine-2.0.3.dist-info/RECORD,,

corbell/__init__.py

@@ -2,5 +2,5 @@
 Corbell — Code retrieval engine for LLM context injection.
 """
 
-__version__ = "2.0.1"
+__version__ = "2.0.3"
 __all__ = ["__version__"]

corbell/cli/commands/debug.py

@@ -0,0 +1,305 @@
+"""CLI: corbell debug — launch a Gradio UI for inspecting query pipeline internals."""
+
+from __future__ import annotations
+
+import os
+from pathlib import Path
+
+import typer
+from rich.console import Console
+
+app = typer.Typer(no_args_is_help=False, help="Query debug UI commands.")
+console = Console()
+
+
+@app.callback(invoke_without_command=True)
+def debug(
+    ctx: typer.Context,
+    workspace: str = typer.Option(
+        "",
+        "--workspace",
+        "-w",
+        help="Path to the workspace root (default: current directory).",
+    ),
+    port: int = typer.Option(7860, "--port", "-p", help="Port for the Gradio server."),
+    share: bool = typer.Option(False, "--share", help="Create a public Gradio share link."),
+) -> None:
+    """Launch the Gradio debug UI for inspecting the query pipeline.
+
+    The UI lets you run a query against a workspace and inspect:
+      - Per-phase timing
+      - Final formatted results
+      - Pre-rerank chunk table (file, lines, score, symbol, type)
+      - LLM rerank prompts and raw response
+    """
+    if ctx.invoked_subcommand is not None:
+        return
+
+    try:
+        import gradio as gr  # type: ignore[import-untyped]
+    except ImportError:
+        console.print(
+            "[red]Gradio is not installed. Install it with:[/red]\n"
+            "  pip install 'codebase-retrieval-context-engine[debug]'"
+        )
+        raise typer.Exit(1)
+
+    default_workspace = workspace or os.environ.get("CORBELL_WORKSPACE") or str(Path.cwd())
+
+    def run_mcp_tool(
+        env_vars_text: str,
+        mcp_workspace: str,
+        mcp_query: str,
+    ):  # type: ignore[no-untyped-def]
+        """Invoke context_engine_codebase_retrieval directly and return results."""
+        if not mcp_query.strip():
+            return "", ""
+
+        # Apply env var overrides for this invocation
+        env_backup: dict[str, str | None] = {}
+        if env_vars_text.strip():
+            for line in env_vars_text.strip().splitlines():
+                line = line.strip()
+                if not line or line.startswith("#"):
+                    continue
+                if "=" not in line:
+                    continue
+                key, _, value = line.partition("=")
+                key, value = key.strip(), value.strip()
+                env_backup[key] = os.environ.get(key)
+                os.environ[key] = value
+
+        try:
+            from corbell.core.mcp.server import context_engine_codebase_retrieval
+
+            result = context_engine_codebase_retrieval(
+                query=mcp_query.strip(),
+                workspace_full_path=mcp_workspace.strip(),
+            )
+
+            if result.startswith("Error:"):
+                return result, ""
+            return "", result
+        except Exception as exc:
+            return f"Error: {exc}", ""
+        finally:
+            for key, original in env_backup.items():
+                if original is None:
+                    os.environ.pop(key, None)
+                else:
+                    os.environ[key] = original
+
+    def run_query(workspace_path: str, query: str):  # type: ignore[no-untyped-def]
+        """Run the debug pipeline and return Gradio component values."""
+        from corbell.core.query.engine import codebase_retrieval_debug
+
+        if not query.strip():
+            return (
+                "",              # error_box
+                "",              # timing_md
+                "",              # final_results
+                [],              # pre_rerank_table
+                "",              # rerank_system
+                "",              # rerank_user
+                "",              # rerank_response
+            )
+
+        ws = workspace_path.strip() or default_workspace
+        result = codebase_retrieval_debug(query=query, workspace_path=ws)
+
+        # --- Error banner ---
+        error_text = result.error or ""
+
+        # --- Timing table ---
+        timing = result.diagnostics.timing if result.diagnostics else {}
+        if timing:
+            rows = "".join(
+                f"| {phase} | {elapsed:.3f}s |\n"
+                for phase, elapsed in timing.items()
+            )
+            timing_md = (
+                "| Phase | Elapsed |\n"
+                "|---|---|\n"
+                + rows
+            )
+        else:
+            timing_md = "_No timing data available._"
+
+        # --- Final results ---
+        final_results = result.final_output or ""
+
+        # --- Pre-rerank table ---
+        pre_rerank_rows = []
+        graph_ids = set()
+        if result.diagnostics and result.diagnostics.graph_chunk_ids:
+            graph_ids = result.diagnostics.graph_chunk_ids
+        for chunk in result.pre_rerank_chunks:
+            chunk_id = getattr(chunk, "chunk_id", "")
+            parts = chunk_id.split("+") if chunk_id else []
+            has_graph = any(p in graph_ids for p in parts) if graph_ids else False
+            has_embedding = any(p not in graph_ids for p in parts) if graph_ids else True
+            if has_graph and has_embedding and len(parts) > 1:
+                source = "embedding+graph"
+            elif has_graph:
+                source = "graph"
+            else:
+                source = "embedding"
+            pre_rerank_rows.append([
+                getattr(chunk, "file_path", ""),
+                f"{getattr(chunk, 'start_line', '')}-{getattr(chunk, 'end_line', '')}",
+                f"{getattr(chunk, 'score', 0.0):.4f}",
+                getattr(chunk, "symbol", "") or "",
+                getattr(chunk, "chunk_type", "") or "",
+                source,
+                getattr(chunk, "content", "") or "",
+            ])
+
+        # --- Rerank prompts ---
+        detail = result.rerank_detail
+        if detail is None or not detail.system_prompt:
+            rerank_system = "_LLM not configured — reranking skipped_"
+            rerank_user = ""
+            rerank_response = ""
+        else:
+            rerank_system = detail.system_prompt
+            rerank_user = detail.user_prompt
+            rerank_response = detail.raw_response or "_No response (LLM call failed)_"
+
+        return (
+            error_text,
+            timing_md,
+            final_results,
+            pre_rerank_rows,
+            rerank_system,
+            rerank_user,
+            rerank_response,
+        )
+
+    with gr.Blocks(title="Corbell Query Debugger") as demo:
+        gr.Markdown("# Corbell Query Debugger")
+        gr.Markdown("Inspect query pipeline internals: timing, pre-rerank chunks, and LLM rerank prompts.")
+
+        with gr.Row():
+            workspace_input = gr.Textbox(
+                label="Workspace Path",
+                value=default_workspace,
+                placeholder="Path to repository root",
+                scale=2,
+            )
+            query_input = gr.Textbox(
+                label="Query",
+                placeholder="e.g. authentication middleware",
+                scale=3,
+            )
+
+        run_btn = gr.Button("Run Query", variant="primary")
+
+        error_box = gr.Textbox(
+            label="Error",
+            visible=True,
+            interactive=False,
+            lines=2,
+        )
+
+        timing_md = gr.Markdown(label="Timing")
+
+        with gr.Tabs():
+            with gr.Tab("Final Results"):
+                final_output = gr.Code(label="Formatted Output", language=None)
+
+            with gr.Tab("Pre-Rerank Chunks"):
+                pre_rerank_table = gr.Dataframe(
+                    headers=["File", "Lines", "Score", "Symbol", "Type", "Source", "Content"],
+                    datatype=["str", "str", "str", "str", "str", "str", "str"],
+                    label="Chunks before reranking",
+                    wrap=False,
+                )
+
+            with gr.Tab("LLM Rerank"):
+                rerank_system_box = gr.Textbox(
+                    label="System Prompt",
+                    lines=6,
+                    interactive=False,
+                )
+                rerank_user_box = gr.Textbox(
+                    label="User Prompt",
+                    lines=12,
+                    interactive=False,
+                )
+                rerank_response_box = gr.Textbox(
+                    label="Raw LLM Response",
+                    lines=4,
+                    interactive=False,
+                )
+
+            with gr.Tab("MCP Debug"):
+                gr.Markdown(
+                    "### MCP Tool Tester\n"
+                    "Configure environment and invoke "
+                    "`context_engine_codebase_retrieval` directly."
+                )
+
+                with gr.Accordion("Environment Configuration", open=False):
+                    mcp_env_vars = gr.Textbox(
+                        label="Environment Variables (one per line, KEY=VALUE)",
+                        placeholder=(
+                            "# Example:\n"
+                            "CORBELL_LLM_PROVIDER=anthropic\n"
+                            "CORBELL_RERANK=true\n"
+                            "ANTHROPIC_API_KEY=sk-..."
+                        ),
+                        lines=6,
+                    )
+
+                gr.Markdown("#### Tool Parameters")
+                with gr.Row():
+                    mcp_workspace_input = gr.Textbox(
+                        label="workspace_full_path",
+                        value=default_workspace,
+                        placeholder="Path to repository root",
+                        scale=3,
+                    )
+                mcp_query_input = gr.Textbox(
+                    label="query",
+                    placeholder="e.g. authentication middleware",
+                )
+
+                mcp_run_btn = gr.Button("Invoke MCP Tool", variant="primary")
+
+                mcp_error_box = gr.Textbox(
+                    label="Error",
+                    visible=True,
+                    interactive=False,
+                    lines=2,
+                )
+                mcp_result_box = gr.Code(
+                    label="Tool Response",
+                    language=None,
+                )
+
+        run_btn.click(
+            fn=run_query,
+            inputs=[workspace_input, query_input],
+            outputs=[
+                error_box,
+                timing_md,
+                final_output,
+                pre_rerank_table,
+                rerank_system_box,
+                rerank_user_box,
+                rerank_response_box,
+            ],
+        )
+
+        mcp_run_btn.click(
+            fn=run_mcp_tool,
+            inputs=[
+                mcp_env_vars,
+                mcp_workspace_input,
+                mcp_query_input,
+            ],
+            outputs=[mcp_error_box, mcp_result_box],
+        )
+
+    console.print(f"[green]Starting Corbell debug UI on port {port}...[/green]")
+    demo.launch(server_port=port, share=share)

corbell/cli/commands/index.py

@@ -2,6 +2,7 @@
 
 from __future__ import annotations
 
+import logging
 import os
 from pathlib import Path
 from typing import Optional
@@ -26,6 +27,9 @@ def build(
     repo: Optional[str] = typer.Option(
         None, "--repo", help="Only index a specific repo by ID."
     ),
+    verbose: bool = typer.Option(
+        False, "--verbose", "-v", help="Enable detailed performance logging."
+    ),
 ) -> None:
     """Build (or incrementally update) the code search index.
 
@@ -39,6 +43,15 @@ def build(
     2. CORBELL_WORKSPACE environment variable
     3. Current working directory
     """
+    if verbose or os.environ.get("CORBELL_VERBOSE", ""):
+        logging.basicConfig(
+            level=logging.INFO,
+            format="%(asctime)s %(name)s %(message)s",
+            datefmt="%H:%M:%S",
+        )
+    else:
+        logging.basicConfig(level=logging.WARNING)
+
     from corbell.core.workspace import build_config, db_path_for_workspace
 
     # Resolve workspace path: flag → env var → cwd

corbell/cli/main.py

@@ -8,6 +8,7 @@ import typer
 from dotenv import load_dotenv
 from rich.console import Console
 
+from corbell.cli.commands.debug import app as debug_app
 from corbell.cli.commands.index import app as index_app
 from corbell.cli.commands.query import app as query_app
 
@@ -32,6 +33,7 @@ console = Console()
 
 app.add_typer(index_app, name="index", help="Code index commands.")
 app.add_typer(query_app, name="query", help="Code search commands.")
+app.add_typer(debug_app, name="debug", help="Query debug UI.")
 
 
 # ---------------------------------------------------------------------------

corbell/core/constants.py

@@ -45,6 +45,14 @@ EXTENSION_LANG: dict[str, str] = {
     ".php": "php",
     ".cs": "csharp",
     ".rs": "rust",
+    ".c": "c",
+    ".cc": "cpp",
+    ".cpp": "cpp",
+    ".cxx": "cpp",
+    ".h": "c",
+    ".hh": "cpp",
+    ".hpp": "cpp",
+    ".hxx": "cpp",
     ".md": "markdown",
     ".yml": "yaml",
     ".yaml": "yaml",

corbell/core/embeddings/extractor.py

@@ -83,7 +83,10 @@ class CodeChunkExtractor:
             lang = _SUPPORTED.get(fp.suffix)
             if not lang:
                 continue
-            rel = str(fp.relative_to(repo_path))
+            rel_path = fp.relative_to(repo_path)
+            if any(part.startswith(".") for part in rel_path.parts[:-1]):
+                continue
+            rel = str(rel_path)
             if gitignore_spec.match_file(rel.replace("\\", "/")):
                 continue
             chunks = self._extract_file(fp, rel, lang, service_id, str(repo_path))

corbell/core/embeddings/model.py

@@ -1,4 +1,4 @@
-"""Embedding model interface + SentenceTransformers implementation."""
+"""Embedding model interface + cloud provider implementations (Google, Voyage)."""
 
 from __future__ import annotations
 
@@ -33,32 +33,6 @@ class EmbeddingModel(ABC):
         ...
 
 
-class SentenceTransformerModel(EmbeddingModel):
-    """Wraps ``sentence-transformers`` with lazy loading.
-
-    Uses ``all-MiniLM-L6-v2`` by default (384-dim, fast, no API key).
-    """
-
-    def __init__(self, model_name: str = "all-MiniLM-L6-v2"):
-        self.model_name = model_name
-        self._model = None  # lazy-loaded
-
-    def _get_model(self):
-        if self._model is None:
-            from sentence_transformers import SentenceTransformer
-            self._model = SentenceTransformer(f"sentence-transformers/{self.model_name}")
-        return self._model
-
-    def encode(self, texts: List[str]) -> List[List[float]]:
-        model = self._get_model()
-        vecs = model.encode(texts, show_progress_bar=False)
-        return [v.tolist() for v in vecs]
-
-    @property
-    def dimension(self) -> int:
-        return self._get_model().get_sentence_embedding_dimension()
-
-
 def _is_voyage_rate_limit_error(e: Exception) -> bool:
     """Return True when a Voyage API error is a 429 rate limit."""
     status = getattr(e, "status_code", None)
@@ -130,9 +104,8 @@ class GoogleEmbeddingModel(EmbeddingModel):
         if not self._api_keys:
             raise ValueError(
                 "GOOGLE_API_KEY is not set. "
-                "Set it in your environment or workspace.yaml:\n"
-                "  export GOOGLE_API_KEY=AIza...\n"
-                "Or use a local embedding model (e.g. all-MiniLM-L6-v2) in storage.model."
+                "Set it in your environment or .env file:\n"
+                "  export GOOGLE_API_KEY=AIza..."
             )
         self._key_index: int = random.randrange(len(self._api_keys))
         # kept for backwards-compat with tests that read _api_key directly
@@ -302,9 +275,8 @@ class VoyageEmbeddingModel(EmbeddingModel):
         if not self._api_keys:
             raise ValueError(
                 "VOYAGE_API_KEY is not set. "
-                "Set it in your environment or workspace.yaml:\n"
-                "  export VOYAGE_API_KEY=pa-...\n"
-                "Or use a local embedding model (e.g. all-MiniLM-L6-v2) in storage.model."
+                "Set it in your environment or .env file:\n"
+                "  export VOYAGE_API_KEY=pa-..."
             )
         self._key_index: int = random.randrange(len(self._api_keys))
         # kept for backwards-compat with tests that read _api_key directly
@@ -367,12 +339,14 @@ class VoyageEmbeddingModel(EmbeddingModel):
                 key = self._api_keys[idx]
                 try:
                     vo = voyageai.Client(api_key=key)
-                    result = vo.embed(
-                        batch,
-                        model=self.model_name,
-                        input_type=input_type,
-                        output_dimension=self.dimension,
-                    )
+                    kwargs: dict = {
+                        "model": self.model_name,
+                        "input_type": input_type,
+                    }
+                    import inspect
+                    if "output_dimension" in inspect.signature(vo.embed).parameters:
+                        kwargs["output_dimension"] = self.dimension
+                    result = vo.embed(batch, **kwargs)
                     self._key_index = (idx + 1) % len(self._api_keys)
                     return result.embeddings
                 except Exception as e: