fix(file_manager): output-token truncation guards + append_file tool (clean cherry-pick from #55) by Starlitnightly · Pull Request #64 · aristoteleo/PantheonOS

fix(prompts): add failure recovery protocol and scientific writing gate

authored and

Starlitnightly

committed

fix(file_manager): add output-token truncation guards and append_file tool

authored and

Starlitnightly

committed

test: add comprehensive tests for output-token truncation guards

Starlitnightly

committed

fix(llm): set max_tokens to model's max output + raise tool guard thresholds

Starlitnightly

committed

fix(llm): adapt max_tokens auto-detection for litellm-free architecture

Starlitnightly

committed

refactor(file_manager): remove size guards, merge append_file into write_file

Starlitnightly

committed

fix(prompts): remove Two-Phase Protocol reference from delegation failure recovery

Starlitnightly

committed

fix(prompts): generalize scientific illustrator description for non-bio domains

Starlitnightly

committed

docs(file_manager): improve write_file docstring with usage guidance

Starlitnightly

committed

refactor(prompts): remove failure recovery section from delegation.md

Starlitnightly

committed

feat(llm): enhance token management and model integration

Starlitnightly

committed

fix(llm): refine output token parameter checks for model responses

Starlitnightly

committed

refactor(openai_adapter): remove unused output token recovery logic

Starlitnightly

committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix(file_manager): output-token truncation guards + append_file tool (clean cherry-pick from #55)#64

fix(file_manager): output-token truncation guards + append_file tool (clean cherry-pick from #55)#64
zqbake merged 13 commits intomainfrom
fix/file-manager-guards

Commits on Apr 4, 2026

fix(prompts): add failure recovery protocol and scientific writing gate

fix(file_manager): add output-token truncation guards and append_file tool

test: add comprehensive tests for output-token truncation guards

fix(llm): set max_tokens to model's max output + raise tool guard thresholds

fix(llm): adapt max_tokens auto-detection for litellm-free architecture

refactor(file_manager): remove size guards, merge append_file into write_file

fix(prompts): remove Two-Phase Protocol reference from delegation failure recovery

fix(prompts): generalize scientific illustrator description for non-bio domains

docs(file_manager): improve write_file docstring with usage guidance

refactor(prompts): remove failure recovery section from delegation.md

feat(llm): enhance token management and model integration

fix(llm): refine output token parameter checks for model responses

refactor(openai_adapter): remove unused output token recovery logic