Nicholas Albion
|
b8fec00b20
|
#131 if `process_name` is specified to `execute_command()` keep the process alive
|
3 years ago |
Nicholas Albion
|
a741261c15
|
`e.msg` may also be 'Unterminated string starting at' or other values
|
3 years ago |
Nicholas Albion
|
86010b3ef3
|
tidy up
|
3 years ago |
Nicholas Albion
|
36f40cf576
|
testing new ApiKeyNotDefinedError in CI
|
3 years ago |
Nicholas Albion
|
7b61d603d5
|
`model = os.getenv('MODEL_NAME', 'gpt-4')` - which makes sense as CI doesn't have my env
|
3 years ago |
Nicholas Albion
|
8247821614
|
more debug
|
3 years ago |
Nicholas Albion
|
8b5951dc71
|
more logging
|
3 years ago |
Nicholas Albion
|
395a08a40f
|
added debuggin logs for CI
|
3 years ago |
Nicholas Albion
|
4a145f7b9b
|
avoid getting stuck in a loop if LLM can't conform to schema.
|
3 years ago |
Nicholas Albion
|
ee7dca9891
|
try without `logger.error(str, e)`
|
3 years ago |
Nicholas Albion
|
718ac2f2ad
|
fix for #130
|
3 years ago |
Nicholas Albion
|
871ac3dcf1
|
`logger.warn()` is deprecated
|
3 years ago |
Nicholas Albion
|
b349702403
|
fixed some unit tests
|
3 years ago |
LeonOstrez
|
f6097d2f7d
|
enable retrying request for extension if API call fails
|
3 years ago |
Nicholas Albion
|
b0094fb7ff
|
added logging
|
3 years ago |
Nicholas Albion
|
f86afce0c6
|
GPT4 responded with a JSON object != schema.
Retry the request, advising what the issue was
|
3 years ago |
LeonOstrez
|
c255a7da10
|
fix updating of app status
|
3 years ago |
LeonOstrez
|
8c3ba579aa
|
fix loading project steps
|
3 years ago |
LeonOstrez
|
dbbfada105
|
fixes
|
3 years ago |
LeonOstrez
|
9adc709de4
|
fix getting args.root, ask feedback only on success and keyboard interrupt (not on exceptions), fix terminal width and erasing old logs
|
3 years ago |
Nicholas Albion
|
b87feefaab
|
log prompt name at INFO
|
3 years ago |
Nicholas Albion
|
3f183b19df
|
Improved logging
|
3 years ago |
pavel-pythagora
|
2929280c04
|
Add arguments for extension
|
3 years ago |
Nicholas Albion
|
3129e34903
|
Merge branch 'main' into debugging_ipc
# Conflicts:
# README.md
# pilot/helpers/agents/CodeMonkey.py
# pilot/helpers/agents/Developer.py
# pilot/prompts/system_messages/architect.prompt
# pilot/utils/arguments.py
# pilot/utils/llm_connection.py
# pilot/utils/utils.py
|
3 years ago |
Nicholas Albion
|
f5d14c1ecc
|
ignore ': OPENROUTER PROCESSING'
|
3 years ago |
Nicholas Albion
|
884bec7531
|
fixed test for CI
|
3 years ago |
Nicholas Albion
|
a4fcd98c95
|
Merge branch 'debugging_ipc' into feature/99-remove-function-calling
|
3 years ago |
Nicholas Albion
|
2cb76e6faf
|
switched to `colorama` which works on Windows and in VS Code
|
3 years ago |
Nicholas Albion
|
2073cfa2ed
|
Improved JSON prompting for GPT-4 and recover incomplete JSON responses from Code Llama
|
3 years ago |
LeonOstrez
|
f65fb84816
|
update code to support python <3.10, update requirements, fix readme issues
|
3 years ago |
Nicholas Albion
|
d9e0df3442
|
removed TODOs
|
3 years ago |
Nicholas Albion
|
0b1fcb5834
|
JSON validation working
|
3 years ago |
Nicholas Albion
|
7013adc28d
|
rejecting responses that are not JSON.
Need to fix prompts for GPT-4
|
3 years ago |
Nicholas Albion
|
03c20af15b
|
notes and testing
|
3 years ago |
Nicholas Albion
|
e1413c82cd
|
removed local_llm_function_calling
|
3 years ago |
Nicholas Albion
|
5b9b0f0da1
|
fixed tests
|
3 years ago |
Nicholas Albion
|
dd995953c0
|
clean up
|
3 years ago |
Nicholas Albion
|
cb3d4b7ad1
|
6 models returning JSON as per ARCHITECTURE function_calls schema
|
3 years ago |
Nicholas Albion
|
0fae65395b
|
anthropic/claude-2 returns JSON
gpt-3_5 returns technologies such as "Backend: Node.js with Mongo database (Mongoose)"
codellama throws an error due to missing `choices`
|
3 years ago |
Nicholas Albion
|
48e57edaba
|
function_call-style JSON response from gpt-4, gpt-3_5, codellama, palm-2-chat-bison
|
3 years ago |
Nicholas Albion
|
376eaaf4e5
|
Refactored AgentConvo.postprecess_response() into function_calling.process_json_response()
|
3 years ago |
Nicholas Albion
|
8603ae7457
|
Tested Developer.install_technology before refactoring without function_calling
|
3 years ago |
Nicholas Albion
|
5853b95ab9
|
ARCHITECTURE function_calls works on meta-llama/codellama-34b-instruct
|
3 years ago |
Nicholas Albion
|
07f9842312
|
tidy up tests
|
3 years ago |
Nicholas Albion
|
886c00a2c6
|
test_chat_completion_Architect passes
|
3 years ago |
Nicholas Albion
|
a32fe03392
|
added `test_end_to_end()` and llm_connection test
|
3 years ago |
Nicholas Albion
|
03e486d899
|
`get_prompt()` moved from `llm_connection` to `utils` and works from unit tests
|
3 years ago |
Nicholas Albion
|
1c568aa9fc
|
fabulous does not work on Windows
|
3 years ago |
Zvonimir Sabljic
|
1b2715b1fe
|
Enabled catching of max token limit errors from OpenAI's response
|
3 years ago |
LeonOstrez
|
b0096a2079
|
merge master into debugging_ipc branch
|
3 years ago |