merge upstream#2

Open

francislabountyjr wants to merge 711 commits intoserp-ai:mainfrom

This pull request is big! We're only showing the most recent 250 commits

Commits on Jun 18, 2025

Commits on Jun 19, 2025

Commits on Jun 22, 2025

Commits on Jun 26, 2025

Commits on Jun 30, 2025

Fix error with additionalProperties for the OpenAI model

RobinPicard
authored and
rlouf
committed

Commits on Jul 1, 2025

Commits on Jul 3, 2025

Commits on Jul 4, 2025

Fix error with the tokonzer inferenace arg in Transformers

RobinPicard
authored and
rlouf
committed

Commits on Jul 5, 2025

Create the inputs module and modify the treatment of multimodal inputs (#1677)
RobinPicard
authored

Commits on Jul 10, 2025

Commits on Jul 11, 2025

Set add_generation_template to True in chat formatting for Transformers

RobinPicard
authored and
rlouf
committed

Commits on Jul 15, 2025

Update index.md

laitifranz
authored and
RobinPicard
committed
Update getting_started.md to fix class import from Transformer library

laitifranz
authored and
RobinPicard
committed

Commits on Jul 19, 2025

Commits on Jul 30, 2025

import fix in base.py

Oni-giri
authored and
rlouf
committed

Commits on Aug 2, 2025

Create our own logits processor for Xgrammar, add support for mlx
RobinPicard
committed

Commits on Aug 4, 2025

Turn tokens into strings before building the Outlines-Core Vocabulary
RobinPicard
committed

Commits on Aug 8, 2025

Commits on Aug 11, 2025

Create AsyncOpenAI model

RobinPicard
authored and
rlouf
committed

Commits on Aug 14, 2025

Fix a bug in OutlinesCoreLogitsProcessor about advancing when in final state
RobinPicard
committed

Commits on Aug 18, 2025

Commits on Aug 23, 2025

Commits on Aug 25, 2025

Commits on Aug 30, 2025

Update vLLM model references and API key usage in documentation

laitifranz
authored and
RobinPicard
committed
Fix trailing-whitespace

laitifranz
authored and
RobinPicard
committed

Commits on Sep 2, 2025

Commits on Sep 15, 2025

Update the output_type formatting of the llamacpp model
RobinPicard
committed

Commits on Sep 18, 2025

Fix correct handling of chat multimodal inputs in TransformersMultiModal model and update docs for the usage of Chat in TransformersMultiModal

laitifranz
authored and
RobinPicard
committed
Handle empty assets cases where onyl text is provided to MLLMs

laitifranz
authored and
RobinPicard
committed
Remove deprecated function and update test accordingly since Outlines version is now > 1.2.0

laitifranz
authored and
RobinPicard
committed
Update docstring with OpenAI’s multi-modal chat format support, and update tests to verify new functionality for multimodal chat

laitifranz
authored and
RobinPicard
committed
Clarify docstring for Chat class

laitifranz
authored and
RobinPicard
committed
Enhance asset handling in format_chat_input method to validate asset types and raise errors for unsupported content types in TransformersMultiModalTypeAdapter

laitifranz
authored and
RobinPicard
committed
Update tests for transformers MultiModal to the multimodal chat format

laitifranz
authored and
RobinPicard
committed
Update transformers multimodal docs to reflect new changes with Chat and Batching examples

laitifranz
authored and
RobinPicard
committed
Use consistent string quoting for SUPPORTED_ASSETS to avoid Python f-string errors

laitifranz
authored and
RobinPicard
committed
Add tests for handling empty assets and invalid content types in transformers multimodal type adapter

laitifranz
authored and
RobinPicard
committed
Removing SUPPORTED_ASSETS constant for clarity and to avoid pre-commit errors

laitifranz
authored and
RobinPicard
committed
Apply pre-commit suggestions: trim trailing whitespace

laitifranz
authored and
RobinPicard
committed
Remove or from Chat docstring for better readability

laitifranz
authored and
RobinPicard
committed
Update format_chat_input for Chat class to support list of string + assets and the mm chat template format

laitifranz
authored and
RobinPicard
committed
Reintroduce test for list of string+assets

laitifranz
authored and
RobinPicard
committed
Reintroduce test for list of string+assets and add new tests for missing asset and type keys

laitifranz
authored and
RobinPicard
committed
Add example for using list of string + assets with Chat class

laitifranz
authored and
RobinPicard
committed
Add tests for invalid content types and for unsupported asset types

laitifranz
authored and
RobinPicard
committed
Migrate multimodal type adapter tests to use AutoProcessor for a MM scenario, and added Audio and Video tests

laitifranz
authored and
RobinPicard
committed
Added validation for items with multiple keys

laitifranz
authored and
RobinPicard
committed
Add tests for mutiple assets usage in multimodal chat template format

laitifranz
authored and
RobinPicard
committed
Update text input prompt to align with author's preference

laitifranz
authored and
RobinPicard
committed

Commits on Sep 26, 2025

Add batch generation to the MLXLM model
RobinPicard
committed

Commits on Oct 2, 2025

Add Mistral AI integration (models, tests, docs)

yasteven
authored and
RobinPicard
committed
Another pass to address the issues rasied in PR 1755

yasteven
authored and
RobinPicard
committed
Make cosmethic changes to the Mistral model
RobinPicard
committed

Commits on Oct 6, 2025

Add design partner CTA
rlouf
authored

Commits on Oct 14, 2025

Add the device_dtype init parameter to the transformers model
RobinPicard
committed

Commits on Oct 15, 2025

Commits on Oct 25, 2025

Fix error in the batch method of the MLXLM model
RobinPicard
committed

Commits on Nov 23, 2025

fix: Refactor sampling parameters in VLLMOffline to use StructuredOutputsParams

Ki-Seki
authored and
RobinPicard
committed

Commits on Nov 24, 2025

Add SamplingParams to model response example

Ki-Seki
authored and
RobinPicard
committed
fix: Add SamplingParams to model response examples

Ki-Seki
authored and
RobinPicard
committed
Draft
RobinPicard
committed

Commits on Nov 28, 2025

Use positional unpacking for mlx_lm.load() tuple

Anri-Lombard
authored and
RobinPicard
committed

Commits on Dec 2, 2025

Remove todos

Anri-Lombard
authored and
RobinPicard
committed

Commits on Dec 3, 2025

fix: correct chat format usage

Ki-Seki
authored and
RobinPicard
committed

Commits on Dec 5, 2025

Add AsyncOllama to __all__ exports

Anri-Lombard
authored and
RobinPicard
committed

Commits on Dec 12, 2025

feat: add chat template checking func

Ki-Seki
authored and
RobinPicard
committed
fix: Best-effort use of chat completion for VLLMOffline

Ki-Seki
authored and
RobinPicard
committed
fix: Best-effort use of chat completion for Transformers

Ki-Seki
authored and
RobinPicard
committed
fix: Best-effort use of chat completion for mlxlm

Ki-Seki
authored and
RobinPicard
committed
fix: Best-effort use of chat completion for LlamaCpp

Ki-Seki
authored and
RobinPicard
committed
test: Add some initial tests

Ki-Seki
authored and
RobinPicard
committed
fix: update stream tests to handle potential empty chunks

Ki-Seki
authored and
RobinPicard
committed
fix: add chat_mode parameter to LlamaCpp and from_llamacpp for better chat handling

Ki-Seki
authored and
RobinPicard
committed
fix: remove TODO comments for adding tests in LlamaCpp class methods

Ki-Seki
authored and
RobinPicard
committed
fix: improve error handling for unexpected prompt types and update stream test assertions

Ki-Seki
authored and
RobinPicard
committed
test: simplify some tests

Ki-Seki
authored and
RobinPicard
committed
fix: add tests for _check_hf_chat_template function in tokenizer module

Ki-Seki
authored and
RobinPicard
committed
fix: remove unused AnyTokenizer import from vllm_offline.py

Ki-Seki
authored and
RobinPicard
committed
fix: correct typo

Ki-Seki
authored and
RobinPicard
committed
fix: move tokenizer imports to the appropriate location in vllm_offline.py so that no runtime errors occur

Ki-Seki
authored and
RobinPicard
committed

Commits on Jan 5, 2026

Fix bug in JsonSchema resulting in the schema validation not running
RobinPicard
committed

Commits on Jan 6, 2026

Commits on Jan 7, 2026

Initial lm studio integration

Anri-Lombard
authored and
RobinPicard
committed
Use lmstudio client and asynclient imports

Anri-Lombard
authored and
RobinPicard
committed
Add dependencies to pyptoject and remove learning/

Anri-Lombard
authored and
RobinPicard
committed
Fix lmstudio Python version constraint and regenerate uv.lock

Anri-Lombard
authored and
RobinPicard
committed
Skip lmstudio during cicd

Anri-Lombard
authored and
RobinPicard
committed
Add python version constraint to lmstudio test dependency

Anri-Lombard
authored and
RobinPicard
committed
add lmstudio to omited models

Anri-Lombard
authored and
RobinPicard
committed
Follow ollama test pattern

Anri-Lombard
authored and
RobinPicard
committed
Reset uv.lock to main (maintainers should regenerate with lmstudio)

Anri-Lombard
authored and
RobinPicard
committed
Add multimodal support

Anri-Lombard
authored and
RobinPicard
committed
Add .worktrees to gitignore

Anri-Lombard
authored and
RobinPicard
committed
Add validation and tests

Anri-Lombard
authored and
RobinPicard
committed
Align tests

Anri-Lombard
authored and
RobinPicard
committed
Mock tests for LMStudio
RobinPicard
committed
Update the uv.lock to add lmstudio
RobinPicard
committed
Skip the LMStudio vision tests when mocking
RobinPicard
committed

Commits on Jan 14, 2026

Add step in CI to free disk space
RobinPicard
committed

Commits on Jan 20, 2026

Fix typo in vision argument of Ollama
RobinPicard
committed

Commits on Jan 28, 2026

Commits on Feb 2, 2026

Add parametrized Transformers smoke test for tokenizer robustness

kudos07
authored and
RobinPicard
committed
fix: trailing whitespace for pre-commit

kudos07
authored and
RobinPicard
committed

Commits on Feb 6, 2026

Fix a bug in the Transformers tokenizer
RobinPicard
committed

Commits on Feb 12, 2026

Commits on Mar 2, 2026

Commits on Mar 9, 2026

fix: correct XDG_CACHE_HOME path construction

jnMetaCode
authored and
RobinPicard
committed
Fix linting
RobinPicard
committed
fix(llamacpp): correct EOS attention mask and vocab truncation bugs

giulio-leone
authored and
RobinPicard
committed
test: add coverage for n > size vocab truncation path

giulio-leone
authored and
RobinPicard
committed
fix(llamacpp): guard against n < 0 error returns from llama_token_to_piece

giulio-leone
authored and
RobinPicard
committed
Simplify redundant assertion in test_negative_n_skips_invalid_token

giulio-leone
authored and
RobinPicard
committed

Commits on Mar 17, 2026

fix(dsl): add JSON double-quotes to string literals inside containers

giulio-leone
authored and
RobinPicard
committed
test: add end-to-end regex tests for JSON quoting in containers

giulio-leone
authored and
RobinPicard
committed
test(dsl): satisfy ruff import order

giulio-leone
authored and
RobinPicard
committed
refactor: use String(f'"...") instead of Sequence for JSON quoting

giulio-leone
authored and
RobinPicard
committed
test: update assertions to match refactored _ensure_json_quoted output

giulio-leone
authored and
RobinPicard
committed
fix: accumulate duplicate token IDs in vocabulary construction

RoloPineda
authored and
RobinPicard
committed
style: fix ruff formatting

RoloPineda
authored and
RobinPicard
committed
test: add test for duplicate token ID preservation in vocabulary

RoloPineda
authored and
RobinPicard
committed
fix: add type annotation for mypy

RoloPineda
authored and
RobinPicard
committed
fix: restore type: ignore comments displaced by formatter

RoloPineda
authored and
RobinPicard
committed

Commits on Mar 21, 2026

fix: inline SPIECE_UNDERLINE fallback for transformers compat

giulio-leone
authored and
RobinPicard
committed
test: add coverage for SPIECE_UNDERLINE ImportError fallback

giulio-leone
authored and
RobinPicard
committed
fix(tokenizers): inline spiece underline constant

authored and
RobinPicard
committed

Commits on Mar 26, 2026

Commits on Apr 15, 2026