vexor 0.20.0__py3-none-any.whl → 0.21.0__py3-none-any.whl

vexor/services/index_service.py

@@ -2,7 +2,9 @@
 
 from __future__ import annotations
 
+import itertools
 import os
+from concurrent.futures import ProcessPoolExecutor, ThreadPoolExecutor
 from dataclasses import dataclass, field
 from datetime import datetime, timezone
 from enum import Enum
@@ -15,12 +17,18 @@ from .cache_service import load_index_metadata_safe
 from .content_extract_service import TEXT_EXTENSIONS
 from .js_parser import JSTS_EXTENSIONS
 from ..cache import CACHE_VERSION, IndexedChunk, backfill_chunk_lines
-from ..config import DEFAULT_EMBED_CONCURRENCY
+from ..config import (
+    DEFAULT_EMBED_CONCURRENCY,
+    DEFAULT_EXTRACT_BACKEND,
+    DEFAULT_EXTRACT_CONCURRENCY,
+)
 from ..modes import get_strategy, ModePayload
 
 INCREMENTAL_CHANGE_THRESHOLD = 0.5
 MTIME_TOLERANCE = 5e-1
 MARKDOWN_EXTENSIONS = {".md", ".markdown", ".mdx"}
+_EXTRACT_PROCESS_MIN_FILES = 16
+_CPU_HEAVY_MODES = {"auto", "code", "outline", "full"}
 
 
 class IndexStatus(str, Enum):
@@ -36,6 +44,85 @@ class IndexResult:
     files_indexed: int = 0
 
 
+def _resolve_extract_concurrency(value: int) -> int:
+    return max(int(value or 1), 1)
+
+
+def _resolve_extract_backend(
+    value: str | None,
+    *,
+    mode: str,
+    file_count: int,
+    concurrency: int,
+) -> str:
+    normalized = (value or DEFAULT_EXTRACT_BACKEND).strip().lower()
+    if normalized not in {"auto", "thread", "process"}:
+        normalized = DEFAULT_EXTRACT_BACKEND
+    if normalized == "auto":
+        if (
+            concurrency > 1
+            and file_count >= _EXTRACT_PROCESS_MIN_FILES
+            and mode in _CPU_HEAVY_MODES
+        ):
+            return "process"
+        return "thread"
+    return normalized
+
+
+def _extract_payloads_for_mode(path: Path, mode: str) -> list[ModePayload]:
+    strategy = get_strategy(mode)
+    return strategy.payloads_for_files([path])
+
+
+def _payloads_for_files(
+    strategy,
+    files: Sequence[Path],
+    *,
+    mode: str,
+    extract_concurrency: int,
+    extract_backend: str,
+) -> list[ModePayload]:
+    if not files:
+        return []
+    concurrency = _resolve_extract_concurrency(extract_concurrency)
+    if concurrency <= 1 or len(files) <= 1:
+        return strategy.payloads_for_files(files)
+    max_workers = min(concurrency, len(files))
+
+    def _extract_with_thread_pool() -> list[ModePayload]:
+        def _extract_one(path: Path) -> list[ModePayload]:
+            return strategy.payloads_for_files([path])
+
+        with ThreadPoolExecutor(max_workers=max_workers) as executor:
+            results = executor.map(_extract_one, files)
+            payloads: list[ModePayload] = []
+            for batch in results:
+                payloads.extend(batch)
+            return payloads
+
+    effective_backend = _resolve_extract_backend(
+        extract_backend,
+        mode=mode,
+        file_count=len(files),
+        concurrency=concurrency,
+    )
+    if effective_backend == "process":
+        try:
+            with ProcessPoolExecutor(max_workers=max_workers) as executor:
+                results = executor.map(
+                    _extract_payloads_for_mode,
+                    files,
+                    itertools.repeat(mode),
+                )
+                payloads: list[ModePayload] = []
+                for batch in results:
+                    payloads.extend(batch)
+                return payloads
+        except Exception:
+            return _extract_with_thread_pool()
+    return _extract_with_thread_pool()
+
+
 def build_index(
     directory: Path,
     *,
@@ -46,6 +133,8 @@ def build_index(
     model_name: str,
     batch_size: int,
     embed_concurrency: int = DEFAULT_EMBED_CONCURRENCY,
+    extract_concurrency: int = DEFAULT_EXTRACT_CONCURRENCY,
+    extract_backend: str = DEFAULT_EXTRACT_BACKEND,
     provider: str,
     base_url: str | None,
     api_key: str | None,
@@ -71,6 +160,7 @@ def build_index(
     if not files:
         return IndexResult(status=IndexStatus.EMPTY)
     stat_cache: dict[Path, os.stat_result] = {}
+    extract_concurrency = _resolve_extract_concurrency(extract_concurrency)
 
     existing_meta = load_index_metadata_safe(
         directory,
@@ -111,6 +201,9 @@ def build_index(
                     files=files,
                     missing_rel_paths=missing_line_files,
                     root=directory,
+                    extract_concurrency=extract_concurrency,
+                    extract_backend=extract_backend,
+                    mode=mode,
                 )
                 cache_path = backfill_chunk_lines(
                     root=directory,
@@ -169,7 +262,15 @@ def build_index(
                 path for rel, path in files_with_rel if rel in changed_rel_paths
             ]
             changed_payloads = (
-                strategy.payloads_for_files(changed_files) if changed_files else []
+                _payloads_for_files(
+                    strategy,
+                    changed_files,
+                    mode=mode,
+                    extract_concurrency=extract_concurrency,
+                    extract_backend=extract_backend,
+                )
+                if changed_files
+                else []
             )
 
             cache_path = _apply_incremental_update(
@@ -199,6 +300,9 @@ def build_index(
                     files=files,
                     missing_rel_paths=line_backfill_targets,
                     root=directory,
+                    extract_concurrency=extract_concurrency,
+                    extract_backend=extract_backend,
+                    mode=mode,
                 )
                 cache_path = backfill_chunk_lines(
                     root=directory,
@@ -217,7 +321,13 @@ def build_index(
                 files_indexed=len(files),
             )
 
-    payloads = strategy.payloads_for_files(files)
+    payloads = _payloads_for_files(
+        strategy,
+        files,
+        mode=mode,
+        extract_concurrency=extract_concurrency,
+        extract_backend=extract_backend,
+    )
     file_labels = [payload.label for payload in payloads]
     embeddings = _embed_labels_with_cache(
         searcher=searcher,
@@ -255,6 +365,8 @@ def build_index_in_memory(
     model_name: str,
     batch_size: int,
     embed_concurrency: int = DEFAULT_EMBED_CONCURRENCY,
+    extract_concurrency: int = DEFAULT_EXTRACT_CONCURRENCY,
+    extract_backend: str = DEFAULT_EXTRACT_BACKEND,
     provider: str,
     base_url: str | None,
     api_key: str | None,
@@ -307,7 +419,13 @@ def build_index_in_memory(
         api_key=api_key,
         local_cuda=local_cuda,
     )
-    payloads = strategy.payloads_for_files(files)
+    payloads = _payloads_for_files(
+        strategy,
+        files,
+        mode=mode,
+        extract_concurrency=extract_concurrency,
+        extract_backend=extract_backend,
+    )
     if not payloads:
         empty = np.empty((0, 0), dtype=np.float32)
         metadata = {
@@ -809,6 +927,9 @@ def _build_line_backfill_updates(
     files: Sequence[Path],
     missing_rel_paths: set[str],
     root: Path,
+    extract_concurrency: int,
+    extract_backend: str,
+    mode: str,
 ) -> list[tuple[str, int, int | None, int | None]]:
     if not missing_rel_paths:
         return []
@@ -816,7 +937,13 @@ def _build_line_backfill_updates(
     targets = [files_by_rel[rel] for rel in missing_rel_paths if rel in files_by_rel]
     if not targets:
         return []
-    payloads = strategy.payloads_for_files(targets)
+    payloads = _payloads_for_files(
+        strategy,
+        targets,
+        mode=mode,
+        extract_concurrency=extract_concurrency,
+        extract_backend=extract_backend,
+    )
     return [
         (
             _relative_to_root(payload.file, root),

vexor/services/search_service.py

@@ -7,12 +7,15 @@ from functools import lru_cache
 from pathlib import Path
 import json
 import re
+import numpy as np
 from typing import Sequence, TYPE_CHECKING
 from urllib import error as urlerror
 from urllib import request as urlrequest
 
 from ..config import (
     DEFAULT_EMBED_CONCURRENCY,
+    DEFAULT_EXTRACT_BACKEND,
+    DEFAULT_EXTRACT_CONCURRENCY,
     DEFAULT_FLASHRANK_MAX_LENGTH,
     DEFAULT_FLASHRANK_MODEL,
     DEFAULT_RERANK,
@@ -48,6 +51,8 @@ class SearchRequest:
     temporary_index: bool = False
     no_cache: bool = False
     embed_concurrency: int = DEFAULT_EMBED_CONCURRENCY
+    extract_concurrency: int = DEFAULT_EXTRACT_CONCURRENCY
+    extract_backend: str = DEFAULT_EXTRACT_BACKEND
     rerank: str = DEFAULT_RERANK
     flashrank_model: str | None = None
     remote_rerank: RemoteRerankConfig | None = None
@@ -112,6 +117,15 @@ def _resolve_rerank_candidates(top_k: int) -> int:
     return max(20, min(candidate, 150))
 
 
+def _top_indices(scores: np.ndarray, limit: int) -> list[int]:
+    if limit <= 0:
+        return []
+    if limit >= scores.size:
+        return sorted(range(scores.size), key=lambda idx: (-scores[idx], idx))
+    indices = np.argpartition(-scores, limit - 1)[:limit]
+    return sorted(indices.tolist(), key=lambda idx: (-scores[idx], idx))
+
+
 def _bm25_scores(
     query_tokens: Sequence[str],
     documents: Sequence[Sequence[str]],
@@ -349,6 +363,7 @@ def perform_search(request: SearchRequest) -> SearchResponse:
     from ..cache import (  # local import
         embedding_cache_key,
         list_cache_entries,
+        load_chunk_metadata,
         load_embedding_cache,
         load_index_vectors,
         load_query_vector,
@@ -385,6 +400,8 @@ def perform_search(request: SearchRequest) -> SearchResponse:
             model_name=request.model_name,
             batch_size=request.batch_size,
             embed_concurrency=request.embed_concurrency,
+            extract_concurrency=request.extract_concurrency,
+            extract_backend=request.extract_backend,
             provider=request.provider,
             base_url=request.base_url,
             api_key=request.api_key,
@@ -446,6 +463,7 @@ def perform_search(request: SearchRequest) -> SearchResponse:
 
     file_snapshot = metadata.get("files", [])
     chunk_entries = metadata.get("chunks", [])
+    chunk_ids = metadata.get("chunk_ids", [])
     stale = bool(file_snapshot) and not is_cache_current(
         request.directory,
         request.include_hidden,
@@ -466,6 +484,8 @@ def perform_search(request: SearchRequest) -> SearchResponse:
             model_name=request.model_name,
             batch_size=request.batch_size,
             embed_concurrency=request.embed_concurrency,
+            extract_concurrency=request.extract_concurrency,
+            extract_backend=request.extract_backend,
             provider=request.provider,
             base_url=request.base_url,
             api_key=request.api_key,
@@ -541,7 +561,6 @@ def perform_search(request: SearchRequest) -> SearchResponse:
             index_empty=True,
         )
 
-    from sklearn.metrics.pairwise import cosine_similarity  # local import
     from ..search import SearchResult, VexorSearcher  # local import
     searcher = VexorSearcher(
         model_name=request.model_name,
@@ -595,13 +614,37 @@ def perform_search(request: SearchRequest) -> SearchResponse:
             store_query_vector(int(index_id), query_hash, request.query, query_vector)
         except Exception:  # pragma: no cover - best-effort cache storage
             pass
-    similarities = cosine_similarity(
-        query_vector.reshape(1, -1),
-        file_vectors,
-    )[0]
-    scored = []
-    for idx, (path, score) in enumerate(zip(paths, similarities)):
-        chunk_meta = chunk_entries[idx] if idx < len(chunk_entries) else {}
+    reranker = None
+    rerank = (request.rerank or DEFAULT_RERANK).strip().lower()
+    use_rerank = rerank in {"bm25", "flashrank", "remote"}
+    if use_rerank:
+        candidate_limit = _resolve_rerank_candidates(request.top_k)
+    else:
+        candidate_limit = request.top_k
+    candidate_count = min(len(paths), candidate_limit)
+
+    query_vector = np.asarray(query_vector, dtype=np.float32).ravel()
+    similarities = np.asarray(file_vectors @ query_vector, dtype=np.float32)
+    top_indices = _top_indices(similarities, candidate_count)
+    chunk_meta_by_id: dict[int, dict] = {}
+    if chunk_ids:
+        candidate_ids = [
+            chunk_ids[idx] for idx in top_indices if idx < len(chunk_ids)
+        ]
+        if candidate_ids:
+            try:
+                chunk_meta_by_id = load_chunk_metadata(candidate_ids)
+            except Exception:  # pragma: no cover - best-effort metadata lookup
+                chunk_meta_by_id = {}
+    scored: list[SearchResult] = []
+    for idx in top_indices:
+        path = paths[idx]
+        score = similarities[idx]
+        chunk_meta = {}
+        if chunk_ids and idx < len(chunk_ids):
+            chunk_meta = chunk_meta_by_id.get(chunk_ids[idx], {})
+        elif idx < len(chunk_entries):
+            chunk_meta = chunk_entries[idx]
         start_line = chunk_meta.get("start_line")
         end_line = chunk_meta.get("end_line")
         scored.append(
@@ -614,12 +657,8 @@ def perform_search(request: SearchRequest) -> SearchResponse:
                 end_line=int(end_line) if end_line is not None else None,
             )
         )
-    scored.sort(key=lambda item: item.score, reverse=True)
-    reranker = None
-    rerank = (request.rerank or DEFAULT_RERANK).strip().lower()
-    if rerank in {"bm25", "flashrank", "remote"}:
-        candidate_count = min(len(scored), _resolve_rerank_candidates(request.top_k))
-        candidates = scored[:candidate_count]
+    if use_rerank:
+        candidates = scored
         if rerank == "bm25":
             candidates = _apply_bm25_rerank(request.query, candidates)
             reranker = "bm25"
@@ -662,6 +701,8 @@ def _perform_search_with_temporary_index(request: SearchRequest) -> SearchRespon
         model_name=request.model_name,
         batch_size=request.batch_size,
         embed_concurrency=request.embed_concurrency,
+        extract_concurrency=request.extract_concurrency,
+        extract_backend=request.extract_backend,
         provider=request.provider,
         base_url=request.base_url,
         api_key=request.api_key,
@@ -680,7 +721,6 @@ def _perform_search_with_temporary_index(request: SearchRequest) -> SearchRespon
             index_empty=True,
         )
 
-    from sklearn.metrics.pairwise import cosine_similarity  # local import
     from ..search import SearchResult, VexorSearcher  # local import
 
     searcher = VexorSearcher(
@@ -717,13 +757,23 @@ def _perform_search_with_temporary_index(request: SearchRequest) -> SearchRespon
                 )
             except Exception:  # pragma: no cover - best-effort cache storage
                 pass
-    similarities = cosine_similarity(
-        query_vector.reshape(1, -1),
-        file_vectors,
-    )[0]
+    reranker = None
+    rerank = (request.rerank or DEFAULT_RERANK).strip().lower()
+    use_rerank = rerank in {"bm25", "flashrank", "remote"}
+    if use_rerank:
+        candidate_limit = _resolve_rerank_candidates(request.top_k)
+    else:
+        candidate_limit = request.top_k
+    candidate_count = min(len(paths), candidate_limit)
+
+    query_vector = np.asarray(query_vector, dtype=np.float32).ravel()
+    similarities = np.asarray(file_vectors @ query_vector, dtype=np.float32)
+    top_indices = _top_indices(similarities, candidate_count)
     chunk_entries = metadata.get("chunks", [])
-    scored = []
-    for idx, (path, score) in enumerate(zip(paths, similarities)):
+    scored: list[SearchResult] = []
+    for idx in top_indices:
+        path = paths[idx]
+        score = similarities[idx]
         chunk_meta = chunk_entries[idx] if idx < len(chunk_entries) else {}
         start_line = chunk_meta.get("start_line")
         end_line = chunk_meta.get("end_line")
@@ -737,12 +787,8 @@ def _perform_search_with_temporary_index(request: SearchRequest) -> SearchRespon
                 end_line=int(end_line) if end_line is not None else None,
             )
         )
-    scored.sort(key=lambda item: item.score, reverse=True)
-    reranker = None
-    rerank = (request.rerank or DEFAULT_RERANK).strip().lower()
-    if rerank in {"bm25", "flashrank", "remote"}:
-        candidate_count = min(len(scored), _resolve_rerank_candidates(request.top_k))
-        candidates = scored[:candidate_count]
+    if use_rerank:
+        candidates = scored
         if rerank == "bm25":
             candidates = _apply_bm25_rerank(request.query, candidates)
             reranker = "bm25"
@@ -908,6 +954,7 @@ def _filter_index_by_extensions(
     ext_set = {ext.lower() for ext in extensions if ext}
     if not ext_set:
         return list(paths), file_vectors, metadata
+    chunk_ids = metadata.get("chunk_ids")
     keep_indices: list[int] = []
     filtered_paths: list[Path] = []
     for idx, path in enumerate(paths):
@@ -922,6 +969,8 @@ def _filter_index_by_extensions(
             ext_set,
         )
         filtered_metadata["chunks"] = []
+        if chunk_ids is not None:
+            filtered_metadata["chunk_ids"] = []
         return [], filtered_vectors, filtered_metadata
     filtered_vectors = file_vectors[keep_indices]
     chunk_entries = metadata.get("chunks", [])
@@ -934,6 +983,10 @@ def _filter_index_by_extensions(
         ext_set,
     )
     filtered_metadata["chunks"] = filtered_chunks
+    if chunk_ids is not None:
+        filtered_metadata["chunk_ids"] = [
+            chunk_ids[idx] for idx in keep_indices if idx < len(chunk_ids)
+        ]
     return filtered_paths, filtered_vectors, filtered_metadata
 
 
@@ -946,6 +999,7 @@ def _filter_index_by_exclude_patterns(
 ) -> tuple[list[Path], Sequence[Sequence[float]], dict]:
     if exclude_spec is None:
         return list(paths), file_vectors, metadata
+    chunk_ids = metadata.get("chunk_ids")
     keep_indices: list[int] = []
     filtered_paths: list[Path] = []
     root_resolved = root.resolve()
@@ -966,6 +1020,8 @@ def _filter_index_by_exclude_patterns(
             exclude_spec,
         )
         filtered_metadata["chunks"] = []
+        if chunk_ids is not None:
+            filtered_metadata["chunk_ids"] = []
         return [], filtered_vectors, filtered_metadata
     filtered_vectors = file_vectors[keep_indices]
     chunk_entries = metadata.get("chunks", [])
@@ -978,6 +1034,10 @@ def _filter_index_by_exclude_patterns(
         exclude_spec,
     )
     filtered_metadata["chunks"] = filtered_chunks
+    if chunk_ids is not None:
+        filtered_metadata["chunk_ids"] = [
+            chunk_ids[idx] for idx in keep_indices if idx < len(chunk_ids)
+        ]
     return filtered_paths, filtered_vectors, filtered_metadata
 
 
@@ -994,6 +1054,7 @@ def _filter_index_by_directory(
         relative_dir = directory.resolve().relative_to(index_root.resolve())
     except ValueError:
         return list(paths), file_vectors, metadata
+    chunk_ids = metadata.get("chunk_ids")
     keep_indices: list[int] = []
     filtered_paths: list[Path] = []
     for idx, path in enumerate(paths):
@@ -1014,6 +1075,8 @@ def _filter_index_by_directory(
             recursive=recursive,
         )
         filtered_metadata["chunks"] = []
+        if chunk_ids is not None:
+            filtered_metadata["chunk_ids"] = []
         filtered_metadata["root"] = str(directory)
         return [], filtered_vectors, filtered_metadata
     filtered_vectors = file_vectors[keep_indices]
@@ -1028,6 +1091,10 @@ def _filter_index_by_directory(
         recursive=recursive,
     )
     filtered_metadata["chunks"] = filtered_chunks
+    if chunk_ids is not None:
+        filtered_metadata["chunk_ids"] = [
+            chunk_ids[idx] for idx in keep_indices if idx < len(chunk_ids)
+        ]
     filtered_metadata["root"] = str(directory)
     return filtered_paths, filtered_vectors, filtered_metadata
 

vexor/text.py

@@ -59,6 +59,8 @@ class Messages:
     HELP_SET_MODEL = "Set the default embedding model."
     HELP_SET_BATCH = "Set the default batch size (0 = single request)."
     HELP_SET_EMBED_CONCURRENCY = "Set the number of concurrent embedding requests."
+    HELP_SET_EXTRACT_CONCURRENCY = "Set the number of concurrent file extraction workers."
+    HELP_SET_EXTRACT_BACKEND = "Set the extraction backend (auto, thread, process)."
     HELP_SET_PROVIDER = "Set the default embedding provider (e.g., gemini, openai, custom, or local)."
     HELP_SET_BASE_URL = "Override the provider's base URL (leave unset for official endpoints)."
     HELP_CLEAR_BASE_URL = "Remove the custom base URL override."
@@ -117,6 +119,10 @@ class Messages:
     ERROR_EMPTY_QUERY = "Query text must not be empty."
     ERROR_BATCH_NEGATIVE = "Batch size must be >= 0"
     ERROR_CONCURRENCY_INVALID = "Embedding concurrency must be >= 1"
+    ERROR_EXTRACT_CONCURRENCY_INVALID = "Extraction concurrency must be >= 1"
+    ERROR_EXTRACT_BACKEND_INVALID = (
+        "Unsupported extraction backend '{value}'. Allowed values: {allowed}."
+    )
     ERROR_MODE_INVALID = "Unsupported mode '{value}'. Allowed values: {allowed}."
     ERROR_PROVIDER_INVALID = "Unsupported provider '{value}'. Allowed values: {allowed}."
     ERROR_RERANK_INVALID = "Unsupported rerank value '{value}'. Allowed values: {allowed}."
@@ -266,6 +272,8 @@ class Messages:
     INFO_MODEL_SET = "Default model set to {value}."
     INFO_BATCH_SET = "Default batch size set to {value}."
     INFO_EMBED_CONCURRENCY_SET = "Embedding concurrency set to {value}."
+    INFO_EXTRACT_CONCURRENCY_SET = "Extraction concurrency set to {value}."
+    INFO_EXTRACT_BACKEND_SET = "Extraction backend set to {value}."
     INFO_PROVIDER_SET = "Default provider set to {value}."
     INFO_BASE_URL_SET = "Base URL override set to {value}."
     INFO_BASE_URL_CLEARED = "Base URL override cleared."
@@ -308,6 +316,8 @@ class Messages:
         "Default model: {model}\n"
         "Default batch size: {batch}\n"
         "Embedding concurrency: {concurrency}\n"
+        "Extract concurrency: {extract_concurrency}\n"
+        "Extract backend: {extract_backend}\n"
         "Auto index: {auto_index}\n"
         "Rerank: {rerank}\n"
         "{flashrank_line}"

{vexor-0.20.0.dist-info → vexor-0.21.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: vexor
-Version: 0.20.0
+Version: 0.21.0
 Summary: A vector-powered CLI for semantic search over files.
 Project-URL: Repository, https://github.com/scarletkc/vexor
 Author: scarletkc
@@ -150,13 +150,26 @@ for hit in response.results:
 By default it reads `~/.vexor/config.json`. For runtime config overrides, cache
 controls, and per-call options, see [`docs/api/python.md`](https://github.com/scarletkc/vexor/tree/main/docs/api/python.md).
 
+## AI Agent Skill
+
+This repo includes a skill for AI agents to use Vexor effectively:
+
+```bash
+vexor install --skills claude  # Claude Code
+vexor install --skills codex   # Codex
+```
+
+Skill source: [`plugins/vexor/skills/vexor-cli`](https://github.com/scarletkc/vexor/raw/refs/heads/main/plugins/vexor/skills/vexor-cli/SKILL.md)
+
 ## Configuration
 
 ```bash
 vexor config --set-provider openai          # default; also supports gemini/custom/local
 vexor config --set-model text-embedding-3-small
 vexor config --set-batch-size 0             # 0 = single request
-vexor config --set-embed-concurrency 2       # parallel embedding requests
+vexor config --set-embed-concurrency 4       # parallel embedding requests
+vexor config --set-extract-concurrency 4     # parallel file extraction workers
+vexor config --set-extract-backend auto      # auto|thread|process (default: auto)
 vexor config --set-auto-index true          # auto-index before search (default)
 vexor config --rerank bm25                  # optional BM25 rerank for top-k results
 vexor config --rerank flashrank             # FlashRank rerank (requires optional extra)
@@ -298,17 +311,6 @@ Re-running `vexor index` only re-embeds changed files; >50% changes trigger full
 
 Porcelain output fields: `rank`, `similarity`, `path`, `chunk_index`, `start_line`, `end_line`, `preview` (line fields are `-` when unavailable).
 
-## AI Agent Skill
-
-This repo includes a skill for AI agents to use Vexor effectively:
-
-```bash
-vexor install --skills claude  # Claude Code
-vexor install --skills codex   # Codex
-```
-
-Skill source: [`plugins/vexor/skills/vexor-cli`](https://github.com/scarletkc/vexor/raw/refs/heads/main/plugins/vexor/skills/vexor-cli/SKILL.md)
-
 ## Documentation
 
 See [docs](https://github.com/scarletkc/vexor/tree/main/docs) for more details.

{vexor-0.20.0.dist-info → vexor-0.21.0.dist-info}/RECORD

@@ -1,33 +1,33 @@
-vexor/__init__.py,sha256=DAhiDVmFBHG2bu_3wkHBS2OubxR3yjbqa1nFLjQ0-Uw,441
+vexor/__init__.py,sha256=i0ly8cFA4N_PEQ_rhYgoLp2NPRQc3_ln8Gfi8QWjXSQ,441
 vexor/__main__.py,sha256=ZFzom1wCfP6TPXe3aoDFpNcUgjbCZ7Quy_vfzNsH5Fw,426
-vexor/api.py,sha256=84GxMt4laq9bpfesPJSFUzkTTCrCpIDDRpVLXKGZ-rg,10470
-vexor/cache.py,sha256=B1seuKU0eYLNvFi7Lpy_X13cSvBfhsuQeH4rZ7Hc29Y,45106
-vexor/cli.py,sha256=hnANtRGO5ypEftMyuTmlZhttSuBZy9ivxymQK11gZ9c,65736
-vexor/config.py,sha256=f5Wom1yUzScp52xpdhLlCG6x7ZEgtFV7kQlarxvD9hU,15372
+vexor/api.py,sha256=YCHpiydbPbRJUqdQYrpwe1JrRI-w_7LRuyZDGBP1_d4,11506
+vexor/cache.py,sha256=3i9FKFLSyZ1kx-w1apc12umPaQxWqMP-P8_lvo67hBw,52832
+vexor/cli.py,sha256=M9GKdD_mJ068Zpm62znTp0KhhKp1dkh_WHmfJHR9hwU,68094
+vexor/config.py,sha256=CiPfEH7Ilt6XepEx4p02qfW5HfkpNDBjhEMyckbSWaA,17413
 vexor/modes.py,sha256=N_wAWoqbxmCfko-v520p59tpAYvUwraCSSQRtMaF4ac,11549
 vexor/output.py,sha256=iooZgLlK8dh7ajJ4XMHUNNx0qyTVtD_OAAwrBx5MeqE,864
 vexor/search.py,sha256=MSU4RmH6waFYOofkIdo8_ElTiz1oNaKuvr-3umif7Bs,6826
-vexor/text.py,sha256=ntqwx2hP5QtlUXnIvBh1NFSn8cxRVhvYtFb7aMt_Tus,24171
+vexor/text.py,sha256=2aK5nJHkosmbmyzp9o_Tzb3YlmVnju_IX8BcEPUdhTA,24794
 vexor/utils.py,sha256=GzfYW2rz1-EuJjkevqZVe8flLRtrQ60OWMmFNbMh62k,12472
 vexor/providers/__init__.py,sha256=kCEoV03TSLKcxDUYVNjXnrVoLU5NpfNXjp1w1Ak2imE,92
-vexor/providers/gemini.py,sha256=-bKSubZRELefJmZzclepXNSWUPsXo94EAM9l0JtfbFM,3739
+vexor/providers/gemini.py,sha256=IWHHjCMJC0hUHQPhuaJ_L_97c_mnOXkPkCVdrIR6z-g,5705
 vexor/providers/local.py,sha256=5X_WYCXgyBGIVvvVLgMnDjTkPR4GBF0ksNPyviBlB7w,4838
-vexor/providers/openai.py,sha256=KylfxVxoTRHrK7KHwMotgD7_fq-CLhx43MQGeGz2dfo,3388
+vexor/providers/openai.py,sha256=YnJDY9gJW7RfGGdkgswVHvmOKNvgLRQUsbpA1MUuLPg,5356
 vexor/services/__init__.py,sha256=dA_i2N03vlYmbZbEK2knzJLWviunkNWbzN2LWPNvMk0,160
 vexor/services/cache_service.py,sha256=ywt6AgupCJ7_wC3je4znCMw5_VBouw3skbDTAt8xw6o,1639
-vexor/services/config_service.py,sha256=yJTBbOmxpbzskHPuLlxYXQ-COJC6-qKtvMsSfuneJoA,4471
+vexor/services/config_service.py,sha256=PojolfbSKh9pW8slF4qxCOs9hz5L6xvjf_nB7vfVlsU,5039
 vexor/services/content_extract_service.py,sha256=zdhLxpNv70BU7irLf3Uc0ou9rKSvdjtrDcHkgRKlMn4,26421
-vexor/services/index_service.py,sha256=pteAG-eRA8FJmDc4GEwhHXGZE8Dm5L8uqzBB0Y8Rrgo,28312
+vexor/services/index_service.py,sha256=FXf1bBoqj4-K1l38ItxHf6Oh7QHVIdNAdVY2kg_Zoq8,32265
 vexor/services/init_service.py,sha256=3D04hylGA9FRQhLHCfR95nMko3vb5MNBcRb9nWWaUE8,26863
 vexor/services/js_parser.py,sha256=eRtW6KlK4JBYDGbyoecHVqLZ0hcx-Cc0kx6bOujHPAQ,16254
 vexor/services/keyword_service.py,sha256=vmke8tII9kTwRDdBaLHBc6Hpy_B3p98L65iGkCQgtMU,2211
-vexor/services/search_service.py,sha256=_3WMzHNV0MCGWFXqwYCQ-XF08aJwF9L4mOGGnmXOATs,36076
+vexor/services/search_service.py,sha256=K7SiAuMA7bGeyPWOHPMKpFFvzzkj5kHWwa3p94NakJs,38663
 vexor/services/skill_service.py,sha256=Rrgt3OMsKPPiXOiRhSNAWjBM9UNz9qmSWQe3uYGzq4M,4863
 vexor/services/system_service.py,sha256=KPlv83v3rTvBiNiH7vrp6tDmt_AqHxuUd-5RI0TfvWs,24638
 vexor/_bundled_skills/vexor-cli/SKILL.md,sha256=m3FlyqgHBdRwyGPEp8PrUS21K0G2jEl88tRvhSPta08,2798
 vexor/_bundled_skills/vexor-cli/references/install-vexor.md,sha256=IUBShLI1mAxugwUIMAJQ5_j6KcaPWfobe0gSd6MWU7w,1245
-vexor-0.20.0.dist-info/METADATA,sha256=RBTym4NL38S6OjijOg5fWPW9VmD7AuLEEqjZtQMkZqA,13331
-vexor-0.20.0.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
-vexor-0.20.0.dist-info/entry_points.txt,sha256=dvxp6Q1R1d6bozR7TwmpdJ0X_v83MkzsLPagGY_lfr0,40
-vexor-0.20.0.dist-info/licenses/LICENSE,sha256=wP7TAKRll1t9LoYGxWS9NikPM_0hCc00LmlLyvQBsL8,1066
-vexor-0.20.0.dist-info/RECORD,,
+vexor-0.21.0.dist-info/METADATA,sha256=Lc5PHY_Ir3F56ILYe6IBlkwhN6gMQGZvf48f7x_uVDg,13494
+vexor-0.21.0.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
+vexor-0.21.0.dist-info/entry_points.txt,sha256=dvxp6Q1R1d6bozR7TwmpdJ0X_v83MkzsLPagGY_lfr0,40
+vexor-0.21.0.dist-info/licenses/LICENSE,sha256=wP7TAKRll1t9LoYGxWS9NikPM_0hCc00LmlLyvQBsL8,1066
+vexor-0.21.0.dist-info/RECORD,,