Commit Graph

800 Commits

Author SHA1 Message Date
AI Christianson 4d14b9747f fix interactive command input 2025-02-12 17:08:37 -05:00
AI Christianson 0c8a4009dc fix bug where completion message was wiped too early 2025-02-12 16:58:39 -05:00
Ariel Frischer 6e8b0f2e42 chore(output.py): remove debug print statements for cleaner code and improved readability 2025-02-12 13:50:37 -08:00
Ariel Frischer 803acc6166 feat(agent_utils.py): convert fallback response to string for prompt concatenation to ensure proper formatting
refactor(fallback_handler.py): change failed_messages from set to list for ordered message handling
refactor(fallback_handler.py): update handle_failure method to accept ToolExecutionError type for better type safety
refactor(fallback_handler.py): implement _reset_on_new_failure method to encapsulate failure reset logic
feat(fallback_handler.py): add construct_prompt_msg_list method to create structured message list for fallback tool calls
2025-02-12 13:39:25 -08:00
AI Christianson a169ed8517 disable put_complete_file_contents; improve prompts; improve status panel output 2025-02-12 16:20:24 -05:00
Ariel Frischer af9f95ceb1 refactor(agent_utils.py): remove the _handle_tool_execution_error function and simplify error handling in run_agent_with_retry
feat(fallback_handler.py): enhance handle_failure method to extract tool name from ToolExecutionError and improve fallback logic
fix(exceptions.py): update ToolExecutionError to include base_message for better error context
feat(output.py): add base_message to ToolExecutionError for improved debugging
chore(tool_configs.py): update get_all_tools function to specify return type
style(logging_config.py): reorder imports for consistency
test(tests): add tests for new error handling and fallback logic in agent_utils and fallback_handler
2025-02-12 13:07:12 -08:00
AI Christianson 905ed2c8fc improve expert model auto detection 2025-02-12 15:55:47 -05:00
AI Christianson e3a705eb9b auto detect openai expert models 2025-02-12 15:40:21 -05:00
AI Christianson c9d7e90312 lower interactive shell history to conserve context; improve prompts 2025-02-12 15:26:48 -05:00
AI Christianson 94f0d96654 lower interactive shell history to conserve context; improve prompts 2025-02-12 15:26:36 -05:00
AI Christianson 6f9ed9562d increase history; fix test 2025-02-12 15:01:02 -05:00
AI Christianson dc079c5d0e improve interactive tty process capture 2025-02-12 14:58:58 -05:00
AI Christianson 7598d42cf9 set temperature param on all initialize_llm calls 2025-02-12 14:15:05 -05:00
AI Christianson d54203b819 fix tests 2025-02-12 13:52:45 -05:00
AI Christianson 087009918b fix tests 2025-02-12 13:50:45 -05:00
AI Christianson a1371fc7e0 support default temp on a per-model basis; show status panel 2025-02-12 13:38:52 -05:00
AI Christianson 264f5025ed Refactor write file tool so it is easier for LLMs to use properly. 2025-02-12 11:51:19 -05:00
Ariel Frischer a7322eaef2 refactor(fallback_handler.py): clean up code by removing unused imports and comments to enhance readability
refactor(fallback_handler.py): extract tool call extraction logic into a separate method for better organization and maintainability
refactor(fallback_handler.py): introduce _parse_tool_arguments method to handle argument parsing, improving code clarity and reusability
2025-02-11 18:38:52 -08:00
Ariel Frischer 67ecf72a6c feat(fallback): implement fallback handler for tool execution errors to enhance error resilience and user experience
refactor(fallback): streamline fallback model selection and invocation process for improved maintainability
fix(config): reduce maximum tool failures from 3 to 2 to tighten error handling thresholds
style(console): improve error message formatting and logging for better clarity and debugging
chore(main): remove redundant fallback tool model handling from main function to simplify configuration management
2025-02-11 18:35:34 -08:00
Ariel Frischer 1388067769 refactor(agent_utils.py): refactor run_agent_with_retry function for better readability and maintainability by extracting helper functions
feat(agent_utils.py): add new helper functions for handling API errors and managing interrupt signals
fix(agent_utils.py): improve error handling in tool execution and retry logic
feat(fallback_handler.py): enhance fallback handling by binding tools correctly during retries
test(tests): add unit tests for new helper functions and refactored logic in agent_utils.py
2025-02-11 12:16:04 -08:00
Ariel Frischer de489584e5 refactor(fallback_handler.py): improve code readability by formatting imports and restructuring for loops
fix(fallback_handler.py): ensure fallback models have a default type of "prompt" and handle exceptions properly during fallback attempts
2025-02-11 01:10:22 -08:00
Ariel Frischer d39be05e39 docs(fallback_handler.py): add detailed docstrings to FallbackHandler methods to improve code documentation and clarity on functionality 2025-02-11 00:44:39 -08:00
Ariel Frischer 3d622911a6 feat(fallback_handler.py): add console notification for tool fallback activation to improve user feedback during failures 2025-02-11 00:40:02 -08:00
Ariel Frischer 0521b3ff9a feat(config.py): add RETRY_FALLBACK_COUNT and RETRY_FALLBACK_DELAY
to configure retry behavior for fallback models

refactor(fallback_handler.py): enhance fallback handling logic to
support both prompt-based and function-calling fallbacks with retries

fix(fallback_handler.py): update fallback model selection to return
dictionaries for better structure and access to model properties
2025-02-11 00:38:15 -08:00
Ariel Frischer 55abf6e5dd feat(fallback_handler): implement FallbackHandler class to manage tool failures and fallback logic
refactor(ciayn_agent): integrate FallbackHandler into CiaynAgent for improved failure handling
fix(agent_utils): add missing newline for better readability in run_agent_with_retry function
test(fallback_handler): add unit tests for FallbackHandler to ensure correct failure handling and fallback logic
2025-02-10 23:37:15 -08:00
Ariel Frischer d8ee4e04f4 feat(fallback): implement automatic fallback to alternative LLM models on consecutive failures to enhance user experience and prevent infinite error loops
refactor(ciayn_agent): restructure tool failure handling to track consecutive failures and fallback attempts more effectively
fix(logging): add pretty logging option for improved log readability
chore(config): define valid providers for LLM selection and update fallback model loading logic
test(ciayn_agent): add unit tests for fallback logic and tool failure handling to ensure reliability and correctness
2025-02-10 14:13:19 -08:00
AI Christianson 589233aaff fix tests 2025-02-10 12:13:14 -05:00
AI Christianson aca62d0c92 update logo 2025-02-10 11:45:36 -05:00
AI Christianson 149e8e2251 set timeouts on llm clients 2025-02-10 11:41:27 -05:00
AI Christianson 26b0f1bdd5 update logo 2025-02-10 11:41:03 -05:00
Ariel Frischer 45b993cfd0 feat(issue): add LLM Tool Call Fallback Feature documentation to outline the new functionality for automatic fallback to alternative LLM models after consecutive failures
feat(ciayn_agent): implement fallback mechanism in CiaynAgent to handle tool call failures and switch to alternative models
feat(__main__): add command line arguments for fallback configuration in the main application
feat(llm): add validation for required environment variables for LLM providers and merge chat histories during fallback
fix(config): define default values for maximum tool failures in configuration
test(ciayn_agent): add unit tests for fallback logic and tool call execution with retries and error handling
test(llm): enhance tests for LLM initialization and environment variable validation
2025-02-09 22:07:23 -08:00
Jose M Leon 00a455d586
FIX do not default to o1 model (#82) 2025-02-08 20:28:10 -05:00
AI Christianson 0c86900ce4 Reduce tool count to make tool calling more reliable. 2025-02-08 18:26:08 -05:00
AI Christianson 13016278e5 prompt improvements 2025-02-08 16:10:24 -05:00
AI Christianson 4c0c2e2ccf prompt improvements 2025-02-08 15:54:18 -05:00
AI Christianson f40e11ee21 improve work logging; use reasoning_effort=high for openai expert models; improve prompts 2025-02-08 14:36:08 -05:00
AI Christianson 5fad3fc755 make cwd/current date available to more agents 2025-02-08 13:58:16 -05:00
AI Christianson ea992960c1 prompt improvements 2025-02-08 13:36:30 -05:00
AI Christianson 2e31bb4ba9 upgrade langgraph deps 2025-02-08 13:34:26 -05:00
AI Christianson c27a75bc26 get rid of pointless truncation message 2025-02-08 12:44:55 -05:00
AI Christianson 5861f3a2bf Adjust token/bytes ratio to resolve errors on swebench-lite. 2025-02-08 08:07:37 -05:00
AI Christianson b673cf61b6 white logo 2025-02-06 19:16:11 -05:00
AI Christianson c59989783d Add python-Levenshtein to deps 2025-02-05 16:17:33 -05:00
AI Christianson 6f865575ee
Update README.md 2025-02-04 15:06:28 -05:00
AI Christianson 04a12b920c README logo 2025-02-04 10:14:48 -05:00
AI Christianson a822ae1332
Update README.md 2025-02-03 15:56:30 -05:00
AI Christianson 0270a9a349 Version bump. 2025-02-02 18:54:27 -05:00
AI Christianson 038f057e59 Fix tests. 2025-02-02 18:53:01 -05:00
AI Christianson 53ccc46392 Add model params for o1 and o3-mini. 2025-02-02 18:48:41 -05:00
Ariel Frischer c14fad6d14
Merge pull request #78 from ariel-frischer/fix-token-limit-bug
Fix token limit bug with custom --research/--planner args
2025-02-01 13:06:37 -08:00