codebase-retrieval-context-engine 2.0.5__py3-none-any.whl → 2.0.6__py3-none-any.whl

{codebase_retrieval_context_engine-2.0.5.dist-info → codebase_retrieval_context_engine-2.0.6.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: codebase-retrieval-context-engine
-Version: 2.0.5
+Version: 2.0.6
 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

{codebase_retrieval_context_engine-2.0.5.dist-info → codebase_retrieval_context_engine-2.0.6.dist-info}/RECORD

@@ -1,4 +1,4 @@
-corbell/__init__.py,sha256=GUZEYd2hejlRYIyIEmvdrYMTDjIT3liSSzQaXItrOho,124
+corbell/__init__.py,sha256=8lQdcrCgCID8TDejlPX3pfWw3rZhnPOMOtaxf-XRMtY,124
 corbell/cli/__init__.py,sha256=5-MP6JIWgp4nDLNIhqP6Gtx97GESaIYg3NGxtRGaMv0,28
 corbell/cli/main.py,sha256=CP5EHizFLaBLF1EohgVo_-XFlm4VaO6peQaSnzyfxAI,1954
 corbell/cli/commands/__init__.py,sha256=0mAOs3RWC7XMZnGRN677hjPCHHQKDq9ASjIr_GQM3js,37
@@ -36,12 +36,12 @@ corbell/core/query/__init__.py,sha256=OCyVRZOyh_eLGhOxR_JYyH6zp8O7qy_-rC3fqGHm7B
 corbell/core/query/diagnostics.py,sha256=o9uIAYFQy8hHua1xLMToSaQPP6xcmnvDJMY3fVg1Dhg,2102
 corbell/core/query/engine.py,sha256=vTFVlXqHavxcR1mIy4KbIRWXx-u_uNHDt4Jb3JRiJ78,18016
 corbell/core/query/enhancer.py,sha256=w5mvm1B8qQZpL6RVhMuhq_rls77hakGSNUyanfkyNEU,3934
-corbell/core/query/formatter.py,sha256=xMr8HE-oxBSEKb514aixY7aoUWGeYoK1w5wnaIlCYEc,2813
+corbell/core/query/formatter.py,sha256=ZtiQwh1DqpDsiILlVbMdxq45Gr1Hf8NgZwa8oL0cSsI,4548
 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=0M8Km2WEO3NX46gT0mF7ma9e0v_HOYXu-t6WgF5U2tI,7262
-codebase_retrieval_context_engine-2.0.5.dist-info/METADATA,sha256=5-myVeI6Z9ecAJFK1FpZV08PfxtcD6VDtr7V7CaxlUk,4036
-codebase_retrieval_context_engine-2.0.5.dist-info/WHEEL,sha256=mffPy8wBnZQn2VnJUU5jE99KsxaSfiyMHV9Yt0aLVxs,87
-codebase_retrieval_context_engine-2.0.5.dist-info/entry_points.txt,sha256=vFB4a4Qb7Ty182usK8deJXiis0UYnGIUDusw0V3Jya8,115
-codebase_retrieval_context_engine-2.0.5.dist-info/licenses/LICENSE,sha256=HrhfyXIkWY2tGFK11kg7vPCqhgh5DcxleloqdhrpyMY,11558
-codebase_retrieval_context_engine-2.0.5.dist-info/RECORD,,
+codebase_retrieval_context_engine-2.0.6.dist-info/METADATA,sha256=WCUALd5QR2cce_KLTB3ag9TKKH3OBgEeRtk7Yj-LWv8,4036
+codebase_retrieval_context_engine-2.0.6.dist-info/WHEEL,sha256=mffPy8wBnZQn2VnJUU5jE99KsxaSfiyMHV9Yt0aLVxs,87
+codebase_retrieval_context_engine-2.0.6.dist-info/entry_points.txt,sha256=vFB4a4Qb7Ty182usK8deJXiis0UYnGIUDusw0V3Jya8,115
+codebase_retrieval_context_engine-2.0.6.dist-info/licenses/LICENSE,sha256=HrhfyXIkWY2tGFK11kg7vPCqhgh5DcxleloqdhrpyMY,11558
+codebase_retrieval_context_engine-2.0.6.dist-info/RECORD,,

corbell/__init__.py

@@ -2,5 +2,5 @@
 Corbell — Code retrieval engine for LLM context injection.
 """
 
-__version__ = "2.0.5"
+__version__ = "2.0.6"
 __all__ = ["__version__"]

corbell/core/query/formatter.py

@@ -12,6 +12,8 @@ if TYPE_CHECKING:
 def format_results(
     chunks: List["ScoredChunk"],
     repo_paths: Dict[str, str],
+    max_output_bytes: int = 80_000,
+    max_line_chars: int = 1000,
 ) -> str:
     """Format scored chunks as annotated code blocks for LLM context injection.
 
@@ -26,16 +28,24 @@ def format_results(
         chunks: Scored chunks to format (pre-sorted by score descending).
         repo_paths: Mapping of repo_id -> absolute repo path string.
                     Used to resolve relative file paths to absolute paths.
+        max_output_bytes: Maximum total output size in bytes. Truncation stops at the
+                          last complete chunk boundary that fits. Defaults to 80 000 (~20K tokens).
+        max_line_chars: Maximum characters per source line before inline truncation.
+                        Defaults to 1000.
 
     Returns:
-        Formatted string with all chunks, separated by blank lines.
+        Formatted string with all chunks, separated by blank lines. If the output
+        exceeds max_output_bytes, a trailing note reports how many results were shown.
     """
     if not chunks:
         return ""
 
+    total = len(chunks)
     blocks: List[str] = []
+    accumulated_bytes = 0
+    truncation_footer = ""
 
-    for chunk in chunks:
+    for n, chunk in enumerate(chunks):
         abs_path = _resolve_absolute_path(chunk.file_path, chunk.repo_id, repo_paths)
 
         # Read the actual lines for this chunk range
@@ -47,16 +57,38 @@ def format_results(
         # Build the header: path#Lstart-end
         header = f"{abs_path}#L{chunk.start_line}-{chunk.end_line}"
 
-        # Build numbered lines
+        # Build numbered lines with per-line truncation
         numbered_lines: List[str] = []
         for i, line in enumerate(lines):
             line_num = chunk.start_line + i
+            if len(line) > max_line_chars:
+                line = line[:max_line_chars] + " [truncated — use Read tool for full content]"
             numbered_lines.append(f"{line_num}: {line}")
 
         block = header + "\n" + "\n".join(numbered_lines)
+
+        # Per-output size gate: check if adding this block would exceed the limit
+        # Account for the separator ("\n\n") between blocks
+        separator_size = 2 if blocks else 0
+        block_bytes = len(block.encode("utf-8"))
+        if accumulated_bytes + separator_size + block_bytes > max_output_bytes:
+            # Collect remaining chunk headers so the agent knows what else is relevant
+            remaining_headers: List[str] = []
+            for remaining in chunks[n:]:
+                rp = _resolve_absolute_path(remaining.file_path, remaining.repo_id, repo_paths)
+                remaining_headers.append(f"{rp}#L{remaining.start_line}-{remaining.end_line}")
+            truncation_footer = (
+                f"\n\n[Showing {n}/{total} results. "
+                f"Remaining (use Read tool):\n"
+                + "\n".join(remaining_headers)
+                + "]"
+            )
+            break
+
         blocks.append(block)
+        accumulated_bytes += separator_size + block_bytes
 
-    return "\n\n".join(blocks)
+    return "\n\n".join(blocks) + truncation_footer
 
 
 def _resolve_absolute_path(