Code

Bedrock token count callbacks#20

Code

Merged

3coins merged 31 commits intolangchain-ai:mainlangchain-ai/langchain-aws:mainfrom NAPTlME:bedrock-token-count-callbacksNAPTlME/langchain-aws:bedrock-token-count-callbacksCopy head branch name to clipboardMay 21, 2024

+268-66

Commits

Commits on Apr 20, 2024

• Updating _prepare_input_and_invoke to use the on_llm_end callback

  Show description for 7a22060
  NAPTlME

  committed
  7a22060

  Copy full SHA for 7a22060

• Consolidated GenerationChunk creation to _stream_response_to_generation_chunk

  Show description for 929e3db
  NAPTlME

  committed
  929e3db

  Copy full SHA for 929e3db

• Added an on_llm_end call to the streaming route

  Show description for 9b840d1
  NAPTlME

  committed
  9b840d1

  Copy full SHA for 9b840d1

Commits on Apr 22, 2024

• Updating async streaming options to match streaming functions wrt messages api and outputs

  NAPTlME

  committed
  5802e53

  Copy full SHA for 5802e53

• Updating _stream_response_to_generation_chunk to get the stop reason from the correct dict

  NAPTlME

  committed
  140333b

  Copy full SHA for 140333b

• Ending on "message_stop" rather that "content_block_stop" for messages api

  Show description for 2f2c8ed
  NAPTlME

  committed
  2f2c8ed

  Copy full SHA for 2f2c8ed

• Updating _prepare_intput_and_invoke to return a packaged llm_output including the stop_reason rather than just usage_info

  Show description for dc9c9cd
  NAPTlME

  committed
  dc9c9cd

  Copy full SHA for dc9c9cd

• Renaming to provider_stop_reason_key_map to better convey the intended use

  NAPTlME

  committed
  a91b3c7

  Copy full SHA for a91b3c7

• _combine_generation_info_for_llm_result is now using the generation_info rather than the full chunks to make it compatible with some calls needed from ChatBedrock

  Show description for 0f345c3
  NAPTlME

  committed
  0f345c3

  Copy full SHA for 0f345c3

• changing to output.get() rather than pop() to prevent the usage from being cleared from the llm_output

  Show description for 25e9caf
  NAPTlME

  committed
  25e9caf

  Copy full SHA for 25e9caf

• Passing generation_info through the ChatGenerationChunk from _stream

  NAPTlME

  committed
  8006cb5

  Copy full SHA for 8006cb5

• Getting llm_output (containing usage and stop_reason from both streaming and non-streaming generation and passing into the ChatResult

  NAPTlME

  committed
  1eeb60f

  Copy full SHA for 1eeb60f

• Linting/Formatting/codespell changes

  NAPTlME

  committed
  afda911

  Copy full SHA for afda911

• Updating a function description

  NAPTlME

  committed
  02893cc

  Copy full SHA for 02893cc

• Line Length for Ruff

  NAPTlME

  committed
  794f9e5

  Copy full SHA for 794f9e5

• Making compatible with python 3.8

  NAPTlME

  committed
  b4c5f61

  Copy full SHA for b4c5f61

Commits on Apr 23, 2024

• Ruff Formatting

  NAPTlME

  committed
  6a870de

  Copy full SHA for 6a870de

• Fixing issue with path and branch specification in lint_diff/format_diff

  Show description for 932204a
  NAPTlME

  committed
  932204a

  Copy full SHA for 932204a

• Addressing linting/formatting issues

  NAPTlME

  committed
  4750b80

  Copy full SHA for 4750b80

Commits on Apr 25, 2024

• Typing compatibility for python < 3.10

  NAPTlME

  committed
  8a2465c

  Copy full SHA for 8a2465c

Commits on Apr 29, 2024

• Merge branch 'langchain-ai:main' into bedrock-token-count-callbacks

  NAPTlME

  authored
  64b3578

  Copy full SHA for 64b3578

Commits on May 3, 2024

• Merge branch 'main' into bedrock-token-count-callbacks

  NAPTlME

  authored
  1526f5a

  Copy full SHA for 1526f5a

Commits on May 17, 2024

• Adding a callback to test token counts on_llm_end

  NAPTlME

  committed
  7a7e39f

  Copy full SHA for 7a7e39f

• Adding integration test to verify token counts and stop reason are captured on_llm_end.

  NAPTlME

  committed
  98d6d09

  Copy full SHA for 98d6d09

• Fixing codespell errors

  NAPTlME

  committed
  0f681cb

  Copy full SHA for 0f681cb

• Merge branch 'langchain-ai:main' into bedrock-token-count-callbacks

  NAPTlME

  authored
  f1350c4

  Copy full SHA for f1350c4

• Updating usage info for input/output tokens to use lists to contain these integer values to prevent errors when merge_dicts is called from langchain_core.utils.

  NAPTlME

  committed
  8908cd8

  Copy full SHA for 8908cd8

• Merge branch 'bedrock-token-count-callbacks' of https://github.com/NAPTlME/langchain-aws into bedrock-token-count-callbacks

  NAPTlME

  committed
  aa33982

  Copy full SHA for aa33982

• Adding the usage_info transformation in which the integers are put in lists to the message_start message type.

  NAPTlME

  committed
  f5e4431

  Copy full SHA for f5e4431

• Moving the usage_info nesting of token counts into lists into a function to reduce the code duplication.

  NAPTlME

  committed
  5eace06

  Copy full SHA for 5eace06

• Ran into an issue where the test_chat_bedrock_streaming_generation_info test would occasionally fail (returned "Hello! How are you doing?" one time)

  NAPTlME

  committed
  bfa0871

  Copy full SHA for bfa0871