Skip to content

Conversation

@pull
Copy link

@pull pull bot commented Nov 25, 2025

See Commits and Changes for more details.


Created by pull[bot] (v2.0.0-alpha.4)

Can you help keep this open source service alive? 💖 Please sponsor : )

zealoushacker and others added 30 commits November 4, 2025 14:30
The workflow runs FROM the public branch (ref: public), so it uses
the workflow file from that branch, not from main. This applies the
git rm --cached fix directly to the public branch where it's needed.
Add comprehensive example showing Claude's Programmatic Tool Calling (PTC)
capabilities for reduced latency and token consumption in agentic workflows.

Includes:
- Mock SkyNet drone delivery API with three tools (get_delivery_logs,
  get_container_logs, check_pod_health)
- Traditional tool calling baseline implementation
- Jupyter notebook demonstrating PTC benefits in production scenarios
- Visualization utilities for API responses

This enables Claude to write code that calls tools programmatically within
the Code Execution environment, eliminating round-trips per invocation.
This notebook demonstrates using a crop tool to improve Claude's performance on chart question answering tasks with the FigureQA dataset.

Key features:
- Baseline evaluation (no tools): 68% accuracy
- With crop tool: 84% accuracy (+16% improvement)
- Tool allows zooming into specific chart regions using normalized coordinates
- Useful for comparing bars, reading labels, and identifying colors
- Add intro explaining both Anthropic SDK and Claude Agent SDK implementations
- Add bonus section demonstrating AgentSDK with custom crop tool
- Fix extract_answer to handle nested JSON with balanced brace matching
- Fix compute_metrics to normalize answers (handle trailing periods)
- Update comparison table to clarify Anthropic SDK vs Claude Agent SDK
…-editing

Update memory cookbook context editing
The answer strings include periods ('yes.'/'no.'), so the counting
logic needed to match that format.
Simplify documentation and examples while preserving key demonstrations.

- Consolidate custom summary prompt example (remove separate execution cell)
- Update token usage statistics and compaction event counts
- Fix code reference: use `len(messages_list)` instead of `len(runner._params['messages'])`
- Simplify list formatting in requirements section
- Remove duplicate best practices section that was shown later
- Clean up and standardize output formatting with visual separators
- Reduce dataset download from full to 10 samples
- Simplify agentic loop (remove max_crops)
- Add Claude Agent SDK section with MCP tool
- Display cropped images inline in output
- Use consistent prompt format across both approaches

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <[email protected]>
Resolved conflict by combining both approaches:
- Kept up-to-date check for efficiency
- Adopted comprehensive workflow exclusion (all workflows, not just sync workflows)
zealoushacker and others added 3 commits November 24, 2025 11:53
Resolved conflicts in:
- .github/workflows/pull-from-public.yml (removed, private repo only)
- pyproject.toml (merged dependencies and ruff config)
- uv.lock (regenerated with uv lock)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <[email protected]>
* Add tool search cookbook

* fixes, run all
@pull pull bot locked and limited conversation to collaborators Nov 25, 2025
@pull pull bot added the ⤵️ pull label Nov 25, 2025
@pull pull bot merged commit bedc03b into code:main Nov 25, 2025
2 of 3 checks passed
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants