Cc - #23
Merged
Merged
Cc#23
Conversation
- M3: Add `code op="usages"` for name-resolved references, extending `tool_code` to include a new operation for querying symbol usages. - M4: Implement `edit` rich descriptor `kind="symbol"` for span-based edits, enhancing the existing `tool_smart_edit` functionality. - M5: Introduce `code op="pattern"` for ast-grep structural search and rewrite, allowing for more precise code refactoring. - M6: Add function-level embeddings and hybrid RRF ranking to improve symbol search results in `CodeContextEngine`. - M7: Implement `memory op="recall_symbol"` to fuse code and memory traces, providing contextual information about symbols. - M8: Add `code op="callers"` and `op="callees"` to traverse the SCIP call graph for a named symbol. - M9: Enable external dependency indexing with `scope="external"` for navigating third-party libraries. - Grounding document updates to reflect the integration of new features and existing MCP tools.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- Introduced M17 milestone for cross-language edge resolution, detailing the goals, module layout, and implementation specifics for resolving common FFI patterns. - Added SQL schema for cross-language edge table and outlined resolver implementations for ctypes, subprocess, and dynamic imports. - Updated existing code operations to include cross-language references in responses. feat: create M18 build-vs-integrate checkpoint documentation - Established M18 milestone to evaluate existing tools against the need for a Zoekt backend. - Defined evaluation criteria for Sourcegraph CLI, self-hosted Sourcegraph, emerging scip-mcp servers, and fallback to Zoekt standalone. - Included a quantitative rubric for candidate evaluation and an output section for decision documentation. fix: update M2 symbol tool scope to include "deleted" - Modified the scope parameter in the M2 symbol tool to support "deleted" symbols, enhancing the functionality for symbol retrieval. fix: clarify external dependencies in M9 documentation - Revised M9 documentation to clarify the behavior of external indexes and the scope parameter, removing the "all" option and specifying the behavior of "external". fix: enhance grounding documentation with cache and budget details - Updated grounding documentation to include details about the shared retrieval cache and budget packer, emphasizing their roles in the code operations. fix: refine index documentation to reflect new milestones - Adjusted index documentation to incorporate new milestones M14, M15, M16, M17, and M18, ensuring clarity on the overall project structure and dependencies.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- assert code search total_tokens stays within budget - capture metadata wrapper overshoot in code_context search Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- add retrieval cache and budget packer exports for code_context - keep packed code payloads within declared token budgets - preserve local and cached provenance metadata on code wrappers Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- use a feasible code search budget threshold for wrapper accounting - keep the regression deterministic across the Phase 1 pytest suite Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- cover code search cache invalidation and budget metadata at MCP boundary - assert uncached and cached code-tool telemetry preserve provenance metadata Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- add a deterministic Phase 1 benchmark runner for repeated code search - cover cache-hit and provenance transitions in the benchmark smoke gate Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- annotate payload builder callbacks for code-context packing helpers - keep the Phase 1 wave gate clean under strict type checking Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- use a feasible MCP search budget threshold for wrapper accounting - keep gateway budget assertions deterministic across targeted regressions Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- record Phase 1 Plan 01 execution summary and verification notes - update state, roadmap, and requirements progress for FNDN-01 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- cover healthy provider routing through the engine delegation seam - assert unhealthy providers fall back to the local symbol path Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- introduce SymbolIntelStore and provider health routing contracts - delegate engine symbol search and symbol lookup through the routed store Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- cover routed search and symbol provenance from synthetic SCIP artifacts - assert artifact refresh invalidates cached routed search results Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- register a repo-local SCIP provider behind the routed symbol store - invalidate retrieval cache state when trusted SCIP artifacts refresh Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- cover routed SCIP provenance, fallback, and cache invalidation at the MCP boundary - enforce the 100x latency floor and 50% token ceiling on deterministic fixtures Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- harden deterministic latency/token benchmarks against routed local-artifact bleed - satisfy lint and mypy checks for the routed SCIP support modules Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- record Phase 1 Plan 02 execution summary and deferred full-suite note - update state, roadmap, and requirements progress for FNDN-02 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- cover additive code search params for snippet scope and glob - expect provenance breakdown and snippet fields on code search Co-authored-by: OpenAI <noreply@openai.com>
- add snippet, scope, and file glob params on code search - attach snippets and provenance breakdown on packed search results Co-authored-by: OpenAI <noreply@openai.com>
- require snippet none to omit snippet fields from code search - catch search budget overflow after snippet defaults were added Co-authored-by: OpenAI <noreply@openai.com>
- make code search default to snippet-free budget-safe responses - drop over-budget trailing hits and omit empty snippet fields Co-authored-by: OpenAI <noreply@openai.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Set default VITE_API_URL to http://atelier-service:8787 so the frontend container can reach the API via Docker network. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- Implemented a benchmark runner in `runner.py` to compare builtin tools against Atelier's MCP across multiple hosts. - Added dataclasses for structured results and reports. - Created functions for benchmarking read, shell, and search operations. - Integrated subprocess calls for MCP and implemented builtin alternatives for comparison. feat(tests): introduce A/B testing for real file benchmarks - Added `test_read_ab_real.py` to measure performance of MCP's read against native file operations. - Included synthetic fixtures for Go, Rust, and Java to ensure comprehensive testing. - Implemented calibration storage for performance metrics in `savings_calibration.jsonl`. - Added tests to validate the correctness of the output and performance savings. chore(fixtures): add sample files for Go, Rust, and Java - Created sample Go, Rust, and Java files to serve as fixtures for benchmarking tests. - These files are patterned after real-world services to ensure realistic testing scenarios.
- Added Antigravity as a supported host in various test files. - Implemented a new stop hook for Codex to emit session summaries. - Updated installation scripts and documentation for Antigravity integration. - Created Antigravity-specific configuration files and templates. - Added new agent instructions and workflow guidance for Antigravity. - Introduced a new command-line script for launching Opencode with session summaries. - Updated tests to reflect changes in host capabilities and configurations.
- Replaced `atelier-status` command with `atelier status` throughout scripts and documentation. - Updated installation scripts to remove references to the `atelier-opencode` wrapper. - Modified status script to check for `atelier` command instead of symlink for `atelier-status`. - Enhanced `status` command to provide a dashboard view of recent runs and auth status. - Removed legacy wrapper installation and uninstallation logic. - Cleaned up tests to reflect changes in command structure and removed tests for obsolete wrappers. - Deleted unused SQLite vector cache file.
- Introduced tests to validate the normalization of legacy strength confidence in trace records. - Added tests to ensure that MCP-only fields are ignored during trace record creation. - Implemented a mock client for OpenMemory to simulate various scenarios in tests. - Enhanced the OpenMemory adapter tests to ensure proper delegation and fallback mechanisms. - Updated the memory store round trip tests to include OpenMemory and verify its behavior. - Refactored existing tests to improve clarity and maintainability.
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
| f"OPENAI_API_KEY={openai_api_key}", | ||
| f"USER={user_id}", | ||
| ] | ||
| api_env.write_text("\n".join(api_lines) + "\n", encoding="utf-8") |
|
|
||
| def verify_audit_bundle(root: Path | str, *, bundle_dir: Path | str) -> dict[str, Any]: | ||
| store_root = Path(root).expanduser().resolve() | ||
| target = Path(bundle_dir).expanduser().resolve() |
|
|
||
| def export_audit_bundle(root: Path | str, *, out_dir: Path | str, since: datetime | None = None) -> dict[str, Any]: | ||
| store_root = Path(root).expanduser().resolve() | ||
| bundle_dir = Path(out_dir).expanduser().resolve() |
…ring sanitization' Co-authored-by: Copilot Autofix powered by AI <62310815+github-advanced-security[bot]@users.noreply.github.com>
…f sensitive information' Co-authored-by: Copilot Autofix powered by AI <62310815+github-advanced-security[bot]@users.noreply.github.com>
pankaj4u4m
added a commit
that referenced
this pull request
Jun 18, 2026
pankaj4u4m
added a commit
that referenced
this pull request
Jun 21, 2026
pankaj4u4m
added a commit
that referenced
this pull request
Jun 21, 2026
pankaj4u4m
added a commit
that referenced
this pull request
Jun 21, 2026
pankaj4u4m
added a commit
that referenced
this pull request
Jun 22, 2026
pankaj4u4m
added a commit
that referenced
this pull request
Jul 2, 2026
pankaj4u4m
added a commit
that referenced
this pull request
Jul 2, 2026
pankaj4u4m
added a commit
that referenced
this pull request
Jul 3, 2026
pankaj4u4m
added a commit
that referenced
this pull request
Jul 8, 2026
Whole-answer anchoring: excerpt-shaped Goods anchor sentence style but not reply length -- a model can obey every sentence rule and still write four such sentences plus a Note. Caveman's level examples are entire answers to a stated question; that framing is what anchors total size. Register pairs now show the question and label each Good 'the complete reply, nothing before or after'. Design doc: third example-trap recorded in recipe #4. Co-Authored-By: atelier <293447754+atelier@users.noreply.github.com>
pankaj4u4m
added a commit
that referenced
this pull request
Jul 9, 2026
Whole-answer anchoring: excerpt-shaped Goods anchor sentence style but not reply length -- a model can obey every sentence rule and still write four such sentences plus a Note. Caveman's level examples are entire answers to a stated question; that framing is what anchors total size. Register pairs now show the question and label each Good 'the complete reply, nothing before or after'. Design doc: third example-trap recorded in recipe #4. Co-Authored-By: atelier <293447754+atelier@users.noreply.github.com>
pankaj4u4m
added a commit
that referenced
this pull request
Jul 10, 2026
Replaces the auto-selected 15s excerpt from the long narrated recording with a dedicated ~18s clip recorded specifically for this GIF. Same crop (border-tight, matches the landing page). Co-authored-by: atelier <293447754+atelier@users.noreply.github.com>
pankaj4u4m
added a commit
that referenced
this pull request
Jul 12, 2026
Replaces the auto-selected 15s excerpt from the long narrated recording with a dedicated ~18s clip recorded specifically for this GIF. Same crop (border-tight, matches the landing page). Co-authored-by: atelier <293447754+atelier@users.noreply.github.com>
pankaj4u4m
added a commit
that referenced
this pull request
Jul 15, 2026
Whole-answer anchoring: excerpt-shaped Goods anchor sentence style but not reply length -- a model can obey every sentence rule and still write four such sentences plus a Note. Caveman's level examples are entire answers to a stated question; that framing is what anchors total size. Register pairs now show the question and label each Good 'the complete reply, nothing before or after'. Design doc: third example-trap recorded in recipe #4. Co-Authored-By: atelier <293447754+atelier@users.noreply.github.com>
pankaj4u4m
added a commit
that referenced
this pull request
Jul 15, 2026
Replaces the auto-selected 15s excerpt from the long narrated recording with a dedicated ~18s clip recorded specifically for this GIF. Same crop (border-tight, matches the landing page). Co-authored-by: atelier <293447754+atelier@users.noreply.github.com>
pankaj4u4m
added a commit
that referenced
this pull request
Jul 24, 2026
Whole-answer anchoring: excerpt-shaped Goods anchor sentence style but not reply length -- a model can obey every sentence rule and still write four such sentences plus a Note. Caveman's level examples are entire answers to a stated question; that framing is what anchors total size. Register pairs now show the question and label each Good 'the complete reply, nothing before or after'. Design doc: third example-trap recorded in recipe #4. Co-Authored-By: atelier <293447754+atelier@users.noreply.github.com>
pankaj4u4m
added a commit
that referenced
this pull request
Jul 24, 2026
Replaces the auto-selected 15s excerpt from the long narrated recording with a dedicated ~18s clip recorded specifically for this GIF. Same crop (border-tight, matches the landing page). Co-authored-by: atelier <293447754+atelier@users.noreply.github.com>
pankaj4u4m
added a commit
that referenced
this pull request
Aug 11, 2026
Whole-answer anchoring: excerpt-shaped Goods anchor sentence style but not reply length -- a model can obey every sentence rule and still write four such sentences plus a Note. Caveman's level examples are entire answers to a stated question; that framing is what anchors total size. Register pairs now show the question and label each Good 'the complete reply, nothing before or after'. Design doc: third example-trap recorded in recipe #4. Co-Authored-By: atelier <293447754+atelier@users.noreply.github.com>
pankaj4u4m
added a commit
that referenced
this pull request
Aug 11, 2026
Replaces the auto-selected 15s excerpt from the long narrated recording with a dedicated ~18s clip recorded specifically for this GIF. Same crop (border-tight, matches the landing page). Co-authored-by: atelier <293447754+atelier@users.noreply.github.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.