v0.1.17.post1

What's Changed

• Override push_to_hub() by @shreyaspimpalgaonkar in #420
• Pass generation params in gemini batch by @kartik4949 in #418
• Kluster.ai backend by @RyanMarten in #422
• feat: add cost processor for litellm and external provider by @kartik4949 in #424
• bump: 0.1.17.post1 by @vutrung96 in #425

Full Changelog: v0.1.17...v0.1.17.post1

v0.1.17

What's Changed

• bump: 0.1.16 by @vutrung96 in #391
• Data generation example by @shreyaspimpalgaonkar in #394
• Update data generation readme by @shreyaspimpalgaonkar in #395
• Rename Bespoke-Stratos example directory by @shreyaspimpalgaonkar in #396
• docs: fix link in citation by @vutrung96 in #397
• ref: block capacity by max_tokens for anthropic online by @adamoptimizer in #387
• Update requirements.txt by @shreyaspimpalgaonkar in #402
• Update README.md by @RyanMarten in #404
• Update README.md by @RyanMarten in #405
• feat: support max parallel request processor by @adamoptimizer in #378
• ref: do not use request file info from batch metadata by @adamoptimizer in #406
• Remove upper bound from tiktoken by @kartik4949 in #409
• Feat/gemini batch processor by @adamoptimizer in #403
• bump: 0.1.17 by @vutrung96 in #417

Full Changelog: v0.1.16...v0.1.17

v0.1.16

What's Changed

• Update README by @RyanMarten in #366
• fix: divide cost by minutes instead of num requests in rate/minute by @vutrung96 in #367
• docs: fix typo in first example by @vutrung96 in #368
• Allow lists of messages as simple input by @RyanMarten in #371
• feat: support separate rate limits for input and output tokens and add moving average estimate of output tokens by @adamoptimizer in #373
• Update CONTRIBUTING.md by @RyanMarten in #382
• Organize dependencies by @RyanMarten in #383
• feat: add support for returning the full completions object + some changes to support deepseek models by @vutrung96 in #384
• fix: check whether a model cost is available before getting cost by @vutrung96 in #385
• ref: make invalid_finish_reasons configurable by @adamoptimizer in #380
• ref: free extra capacity in online processor by @adamoptimizer in #381
• Fix/llm/dataset by @adamoptimizer in #393

Full Changelog: v0.1.15...v0.1.16

v0.1.15.post1

What's Changed

• Update README by @RyanMarten in #366
• fix: divide cost by minutes instead of num requests in rate/minute by @vutrung96 in #367

Full Changelog: v0.1.15...v0.1.15.post1

v0.1.15

What's Changed

• fix: allow special tokens in tiktoken encoding by @devin-ai-integration in #316
• feat: make cache directory configurable via CURATOR_CACHE_DIR environment variable by @devin-ai-integration in #321
• Move prompt_formatter test file to the correct location by @vutrung96 in #323
• Add schema validation to prevent DB schema mismatches by @devin-ai-integration in #322
• Closes #297 Support local models via vLLM by @marianna13 in #298
• fix: update poetry.lock to reflect pyproject.toml changes by @RyanMarten in #326
• Relax VLLM version requirements by @vutrung96 in #329
• Replace black with Ruff and add pre-commit hooks by @shreyaspimpalgaonkar in #302
• Fix circular import by @GeorgiosSmyrnis in #330
• Refactor: Factory Pattern, Base URL via Env, and Backend Determination Fix by @adamoptimizer in #317
• Test/integration/basic setup by @adamoptimizer in #320
• test: overload integration tests with other backends by @adamoptimizer in #334
• Update unittests and add coverage atleast 80 by @adamoptimizer in #337
• fix: set backend to None for SimpleLLM class by default by @vutrung96 in #338
• feat: add ability to disable caching via CURATOR_DISABLE_CACHE by @vutrung96 in #335
• fix: make sure we post process dataset file by sorting by and removing the __original_row_idx column by @vutrung96 in #340
• Rynam/batch retry fix by @adamoptimizer in #343
• Retry failed requests within batches by @RyanMarten in #295
• feat: detailed progress tracking via cli for online request processors by @vutrung96 in #342
• fix: safe open file - closes #344 by @RyanMarten in #346
• test: add anthropic integration tests by @adamoptimizer in #350
• docs: add contribution md by @adamoptimizer in #353
• Update readme and mute camel test by @adamoptimizer in #355
• feat: detailed progress tracking for batch processing by @vutrung96 in #349
• docs: add citation by @RyanMarten in #357
• fix: broken vllm offline processor + add test by @vutrung96 in #358
• ref: refactor LLM params into backend params by @adamoptimizer in #345
• feat: Updated LLM class interface by @vutrung96 in #360
• Add fix_json check by @GeorgiosSmyrnis in #359
• perf: lazy import litellm, datasets by @adamoptimizer in #361
• docs: add Kartik to CITATION.cff by @vutrung96 in #363
• Add venv activation instructions to contributing.md by @RyanMarten in #364

New Contributors

• @marianna13 made their first contribution in #298
• @shreyaspimpalgaonkar made their first contribution in #302
• @adamoptimizer made their first contribution in #317

Full Changelog: v0.1.14...v0.1.15

v0.1.14

What's Changed

• Fix bug in batch mapping and get right order for outputs. by @madiator in #289
• refactor: use os.path.join consistently for path handling by @devin-ai-integration in #291
• Add Anthropic batch and general refactor by @RyanMarten in #243
• Remove duplicate resource limit by @RyanMarten in #299
• Merge dev into main by @vutrung96 in #306
• Re-do docstrings for batch request processors by @devin-ai-integration in #308

Full Changelog: 0.1.13...v0.1.14

0.1.13

What's Changed

1. Fix issues around litellm, to support Gemini Flash Thinking model.
2. Add support for o1.

Details

• Ryan marten patch 1 by @RyanMarten in #273
• Clean ups in llm.py by @madiator in #274
• Put the examples in respective folders and add requirements.txt everywhere by @madiator in #275
• Catch catch-all Exception since litellm doesn't throw specific error. by @madiator in #281
• feat: add o1 model structured output support by @devin-ai-integration in #284
• Bump to 0.1.13 by @madiator in #285
• Merge dev into main for 0.1.13 release. by @madiator in #286

Full Changelog: v0.1.12...0.1.13

v0.1.12

What's Changed

• [curator-viewer] enabled toast instead of alert for copy paste, and fixed streaming toast by @CharlieJCJ in #165
• Use huggingface modified pickler to fix path-dependent caching by @vutrung96 in #230
• Change rpm and tpm to have lower default and allow for manual setting by @RyanMarten in #234
• Various fixes to increase the reliability of batch processing by @vutrung96 in #231
• Graceful error handling for missing requests by @vutrung96 in #244
• OpenAIOnline - if api_key missing, directly error out by @CharlieJCJ in #237
• Increase default values for tpm/rpm, otherwise there is no progress. by @madiator in #245
• refactor: rename Prompter class to LLM by @devin-ai-integration in #242
• Rename prompter. Simplify prompt_formatter and add test. by @madiator in #246
• Raise error on failed responses by @RyanMarten in #251
• Add a SimpleLLM interface, and update documentation. by @madiator in #255
• Cool down when hitting rate limit with online processors by @RyanMarten in #256
• Gemini lower safety constraints by @CharlieJCJ in #259
• Raise on None response message by @RyanMarten in #262
• Add metadata dict + cache verification by @GeorgiosSmyrnis in #257
• Default for all online requests to 10 minutes timeout by @RyanMarten in #265
• Retry only on "max_length" and "content_filter" finish reason by @RyanMarten in #267
• Retry on response format failure by @RyanMarten in #266
• Add prism.js types to dev dependencies by @RyanMarten in #270

New Contributors

• @devin-ai-integration made their first contribution in #242
• @GeorgiosSmyrnis made their first contribution in #257

Full Changelog: v0.1.11...v0.1.12

v0.1.11

What's Changed

• Allow special tokens when encoding text for token accounting by @RyanMarten in #181
• [Package Dependency] Downgrade tiotoken and aiofiles, bump poetry package in pyproject toml by @CharlieJCJ in #183
• Re-merge main into dev by @vutrung96 in #185
• Cleanups and fix minor issues. by @madiator in #184
• Scale batch processor to managing thousands of batches by @RyanMarten in #186
• Easy way to cancel batches by @RyanMarten in #187
• litellm refactoring base online request processor by @CharlieJCJ in #188
• More retries for batch by @RyanMarten in #194
• Delete input and output files for successful batches by @RyanMarten in #195
• Add LiteLLM+instructor (for structured output) backend for curator by @CharlieJCJ in #141
• small bugfix by @RyanMarten in #199
• Increase max retry to 50 by @vutrung96 in #200
• LiteLLM missing cost handling for models by @CharlieJCJ in #210
• OnlineRequestProcessor - Fix retry only once issue by @CharlieJCJ in #202
• Implement persona-hub using Curator by @madiator in #211
• Allow user to switch keys during batch and resume by @RyanMarten in #198
• Small fix for datetime in openai request processor by @CharlieJCJ in #219
• Bump 0.1.11 pypi version by @CharlieJCJ in #221
• 0.1.11 by @CharlieJCJ in #220

Full Changelog: v0.1.10...v0.1.11

v0.1.10

What's Changed

• [curator-viewer] add time logging and curator viewer show distribution by @CharlieJCJ in #149
• enhanced installation UI for curator package by @lavishsaluja in #134
• add cost and token logging in openai online and batching with litellm completion_cost by @CharlieJCJ in #159
• Add lint checks to the repository by @vutrung96 in #168
• Use dill pickle to capture the execution context by @vutrung96 in #167
• 0.1.10 by @CharlieJCJ in #174

Full Changelog: v0.1.9.post1...v0.1.10