Simplify the build#1306

Closed

copybara-service[bot] wants to merge 968 commits intomainfrom

litert_lm_pr_862068350

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

Commits on Nov 8, 2025

[litertlm] simplify sendMessage response handling

whhone
authored and
copybara-github
committed

Commits on Nov 10, 2025

[litertlm] minor improvement for creating Message inline

whhone
authored and
copybara-github
committed

Commits on Nov 11, 2025

Do not prefer textures on Apple devices

tf-marissaw
authored and
copybara-github
committed
[litertlm] Windows JVM build

whhone
authored and
copybara-github
committed

Commits on Nov 12, 2025

[litertlm] Add sendMessageAsync that gets response as Flow

whhone
authored and
copybara-github
committed
Update dependencies of litert_lm

tf-marissaw
authored and
copybara-github
committed
Update dependencies of litert_lm

terryheo
authored and
copybara-github
committed
Update litert_lm with new way of setting GpuOptions

terryheo
authored and
copybara-github
committed

Commits on Nov 13, 2025

internal change and clean up.

ai-edge-bot
authored and
copybara-github
committed
Internal change and clean up.

ai-edge-bot
authored and
copybara-github
committed
Snap the litert dependency to the latest

ai-edge-bot
authored and
copybara-github
committed
litert_lm: Apply recent LiteRt CompiledModel::Create() API change

terryheo
authored and
copybara-github
committed
[litertlm] set the log level to "at least error" in JNI

whhone
authored and
copybara-github
committed

Commits on Nov 15, 2025

Implement `CreateConstraint` in Gemma3DataProcessor. Currently the method returns null.

yishuangP
authored and
copybara-github
committed
Refactor LiteRT LM usages of TensorBuffer creation to use CreateManagedHostMemory when environment is not needed.

gcarranza-1
authored and
copybara-github
committed
[litertlm] update Kotlin API doc

whhone
authored and
copybara-github
committed

Commits on Nov 17, 2025

Internal code change.

ai-edge-bot
authored and
copybara-github
committed
Internal changes ...

yishuangP
authored and
copybara-github
committed
Update dependencies of litert_lm

gcarranza-1
authored and
copybara-github
committed
Remove deprecated TensorBuffer creation methods.

gcarranza-1
authored and
copybara-github
committed
Use TensorBuffer::BufferType that uses C++ type

gcarranza-1
authored and
copybara-github
committed

Commits on Nov 18, 2025

Deprecate SetShloCompositeInlining call in litert lm

ai-edge-bot
authored and
copybara-github
committed
Suppress `fprintf` calls in OSS minja's chat-template.hpp.

yichunk
authored and
copybara-github
committed

Commits on Nov 19, 2025

Add TFlite external weight file support

ai-edge-bot
authored and
copybara-github
committed
Do not reset GPU sampler when llm executor is reset.

yichunk
authored and
copybara-github
committed
Add state tracking to FakeLlmExecutor.

ai-edge-bot
authored and
copybara-github
committed
Rename libLiteRtRuntimeCApi.so to libLiteRt.so

terryheo
authored and
copybara-github
committed
Add a `enable_constrained_decoding` flag to ConversationConfig. Constrained decoding is off by default.

yishuangP
authored and
copybara-github
committed
Update model type inferring logic for coincident BPE error for qwen tokenizer.

yichunk
authored and
copybara-github
committed

Commits on Nov 20, 2025

Update compiled model executor and FakeLlmExecutor to only update constraint decoding state with decoded ids.

ai-edge-bot
authored and
copybara-github
committed
Internal Changes ...

yishuangP
authored and
copybara-github
committed
Create Sampler within ExecutionManager::Create.

ai-edge-bot
authored and
copybara-github
committed
Add LiteRT LM SessionConfig and SamplerParameters to C API.

ai-edge-bot
authored and
copybara-github
committed
Add qwen2p5 llm model type and update data processor factory accordingly.

yichunk
authored and
copybara-github
committed
Automated Code Change

ai-edge-bot
authored and
copybara-github
committed
Update comments in pipeline and executor tests.

ai-edge-bot
authored and
copybara-github
committed
Introduce SessionAdvanced which uses execution manager to execute LLM tasks.

ai-edge-bot
authored and
copybara-github
committed
Do not re-use TensorBuffer with mismatched size in Prefill signature of main llm model.

YunanAZ
authored and
copybara-github
committed

Commits on Nov 21, 2025

Update includes and remove global namespace pollution following styleguide.

ai-edge-bot
authored and
copybara-github
committed
Update litertlm builder for llm model type and jinja template fields in llm metadata

yichunk
authored and
copybara-github
committed
Added MemoryMappedFile to the ModelAsset

ai-edge-bot
authored and
copybara-github
committed

Commits on Nov 22, 2025

Add option to initialize litert_lm model with MemoryMappedFile

ai-edge-bot
authored and
copybara-github
committed
Internal change

hheydary
authored and
copybara-github
committed

Commits on Nov 24, 2025

Commits on Nov 25, 2025

Update the doc with the latest 0.8.0 release

ai-edge-bot
authored and
copybara-github
committed
Add LiteRtLmConversationConfig to C API.

ai-edge-bot
authored and
copybara-github
committed
Add LiteRtLmConversationConfig to C API.

ai-edge-bot
authored and
copybara-github
committed

Commits on Nov 26, 2025

Internal change

matthewchan-g
authored and
copybara-github
committed
Fix comment for GetExecutorSettings.

ai-edge-bot
authored and
copybara-github
committed
Internal code change.

ai-edge-bot
authored and
copybara-github
committed
fix the missing initialized value

ai-edge-bot
authored and
copybara-github
committed
Update litert_lm_conversation_config_create to correctly pass in value for enable_constrained_decoding

ai-edge-bot
authored and
copybara-github
committed

Commits on Dec 1, 2025

Update dependencies of litert_lm

protobird-git
authored and
copybara-github
committed

Commits on Dec 2, 2025

Fix windows build linking capi.dll

protobird-git
authored and
copybara-github
committed
Update dependencies of litert_lm

protobird-git
authored and
copybara-github
committed
[litertlm] internal changes

ai-edge-bot
authored and
copybara-github
committed
Manage session states within execution manager.

ai-edge-bot
authored and
copybara-github
committed
Add a setting to control the use of an external sampler.

ai-edge-bot
authored and
copybara-github
committed

Commits on Dec 3, 2025

[litertlm] version bump

whhone
authored and
copybara-github
committed
Allow trailing commas in Python tool call expressions.

matthewchan-g
authored and
copybara-github
committed

Commits on Dec 4, 2025

[litertlm] fix typo serverity -> severity

whhone
authored and
copybara-github
committed
Internal changes ...

yishuangP
authored and
copybara-github
committed
[litertlm] update tool calling internal json structure

whhone
authored and
copybara-github
committed
Add WaitUntilDone to Engine::Session interface.

ai-edge-bot
authored and
copybara-github
committed

Commits on Dec 5, 2025

Commits on Dec 8, 2025

Internal update ...

yichunk
authored and
copybara-github
committed
Print out the prefill message

ai-edge-bot
authored and
copybara-github
committed
Internal change

matthewchan-g
authored and
copybara-github
committed
Internal change

matthewchan-g
authored and
copybara-github
committed
Introduce engine advanced that creates session_advanced.

ai-edge-bot
authored and
copybara-github
committed
Internal Changes ...

yishuangP
authored and
copybara-github
committed

Commits on Dec 9, 2025

Update last task ID after prefill and add multi-turns for two sessions in main for end to end example.

ai-edge-bot
authored and
copybara-github
committed
Support Windows file paths in file_util

tf-marissaw
authored and
copybara-github
committed
Remove unused includes from litert_lm runtime.

ai-edge-bot
authored and
copybara-github
committed
Add support for chunked prefill in the dynamic LLM executor.

hheydary
authored and
copybara-github
committed
Update dependencies of litert_lm

protobird-git
authored and
copybara-github
committed
Enable Metal Argument Buffers for GPU compilation on Metal backend.

fengwuyao
authored and
copybara-github
committed
Set default input_prompt to empty string since the default prompt string is already returned by the end of GetInputPrompt

ai-edge-bot
authored and
copybara-github
committed
Add option to prefill preface on Conversation initialization.

yichunk
authored and
copybara-github
committed
Internal Build change.

hheydary
authored and
copybara-github
committed
Reverts 29e85bdc11d24ebd39678daa90da5628d01dc6d0

whhone
authored and
copybara-github
committed

Commits on Dec 10, 2025

Run async on GPU by LiteRT-LM

protobird-git
authored and
copybara-github
committed
update the readme doc for a bit more details.

ai-edge-bot
authored and
copybara-github
committed
Add IsLiteRTLMFile function to check for LiteRTLM format.

hheydary
authored and
copybara-github
committed
Ensure tool parsed JSON result always includes a "content" field.

yichunk
authored and
copybara-github
committed
Implement Session Cloning in LiteRT LM.

ai-edge-bot
authored and
copybara-github
committed
Internal changes ...

yishuangP
authored and
copybara-github
committed
Internal code change.

ai-edge-bot
authored and
copybara-github
committed
Internal change

matthewchan-g
authored and
copybara-github
committed

Commits on Dec 12, 2025

Update dependencies of litert_lm

protobird-git
authored and
copybara-github
committed
Set serialization dir properly on Windows

protobird-git
authored and
copybara-github
committed
Add new test .litertlm file with external weights

ai-edge-bot
authored and
copybara-github
committed
In AntlrPythonParser.g4, label dict keys as "key".

matthewchan-g
authored and
copybara-github
committed
A flag to set preferred GPU device.

protobird-git
authored and
copybara-github
committed
Add cxxbridge-cmd, a binary for generating C/C++ bindings to Rust libraries.

matthewchan-g
authored and
copybara-github
committed
Enable the use of Rust protocol buffers in LiteRT-LM OSS.

matthewchan-g
authored and
copybara-github
committed
Internal change

ai-edge-bot
authored and
copybara-github
committed
Update dependencies of litert_lm

protobird-git
authored and
copybara-github
committed
Improve error message for tool call parsing failures.

yishuangP
authored and
copybara-github
committed
Add flags for webgpu threads to upload/compile and gpu conversion

protobird-git
authored and
copybara-github
committed

Commits on Dec 15, 2025

Internal change

ai-edge-bot
authored and
copybara-github
committed

Commits on Dec 16, 2025

Update dependencies of litert_lm

ai-edge-bot
authored and
copybara-github
committed
Add APIs through the model resource to get the ScopedFile and the weights offset.

ai-edge-bot
authored and
copybara-github
committed
Windows + GPU run always need DirectXShaderCompiler

ai-edge-bot
authored and
copybara-github
committed
Enable a hint to LiteRT to reduce memory usage

tf-marissaw
authored and
copybara-github
committed
Internal change

ai-edge-bot
authored and
copybara-github
committed

Commits on Dec 17, 2025

Enable vision and audio modalities in SessionConfig.

ai-edge-bot
authored and
copybara-github
committed

Commits on Dec 18, 2025

Add Function Gemma data processor. (#1123)

copybara-service[bot]
and
matthewchan-g
authored
Add litert_lm_lib test

ai-edge-bot
authored and
copybara-github
committed
Add the capability to enable constrained decoding for Gemma 3/3n models, supporting tool calls in Python-style format.

yishuangP
authored and
copybara-github
committed
Internal code change.

ai-edge-bot
authored and
copybara-github
committed
Add FunctionGemma to the list of supported models and benchmarks.

matthewchan-g
authored and
copybara-github
committed
Update model table in README.md with usage types and FunctionGemma details.

matthewchan-g
authored and
copybara-github
committed
Add the capability to enable constrained decoding for Function Gemma model, supporting tool calls in FC style, Json-based format.

yishuangP
authored and
copybara-github
committed

Commits on Dec 19, 2025

Add tests for multiple tool calls on separate lines when using Python-style tool calls.

matthewchan-g
authored and
copybara-github
committed
Internal update only.

ai-edge-bot
authored and
copybara-github
committed
make sure the litert_lm_lib_test is also added for OSS

ai-edge-bot
authored and
copybara-github
committed
Split GenerateContentStream into Prefill and Async Decode.

ai-edge-bot
authored and
copybara-github
committed
Internal change

ai-edge-bot
authored and
copybara-github
committed
Internal change

ai-edge-bot
authored and
copybara-github
committed
Internal code change.

ai-edge-bot
authored and
copybara-github
committed
Internal change

ai-edge-bot
authored and
copybara-github
committed
Migrate from absl::BitGen to std::default_random_engine for sampling.

ai-edge-bot
authored and
copybara-github
committed
Enable multimodality input clone for advanced session

yichunk
authored and
copybara-github
committed
Add Float16 support to TopPCpuSampler.

fengwuyao
authored and
copybara-github
committed
Internal code change

ai-edge-bot
authored and
copybara-github
committed

Commits on Dec 20, 2025

Updated the engine and session creation logics to enforce that only one session instance can exist for each engine.

ai-edge-bot
authored and
copybara-github
committed
Add Sampler::UpdateConfig method.

ai-edge-bot
authored and
copybara-github
committed
Remove unused sampled_ids_ member.

ai-edge-bot
authored and
copybara-github
committed

Commits on Dec 22, 2025

Update dependencies of litert_lm

protobird-git
authored and
copybara-github
committed
Remove unused variable in io_types.h for chrome's odml import.

ai-edge-bot
authored and
copybara-github
committed
Add single buffer with custom tensor descriptors.

ai-edge-bot
authored and
copybara-github
committed

Commits on Dec 23, 2025

Internal change

ai-edge-bot
authored and
copybara-github
committed

Commits on Dec 24, 2025

Limit avx2 compilation flags only to x86_64 platforms

protobird-git
authored and
copybara-github
committed

Commits on Dec 26, 2025

Use GitHub mirror for BCR to work around bcr.bazel.build outages

ai-edge-bot
authored and
copybara-github
committed

Commits on Dec 29, 2025

Internal change

matthewchan-g
authored and
copybara-github
committed

Commits on Dec 30, 2025

Internal changes only

protobird-git
authored and
copybara-github
committed

Commits on Dec 31, 2025

Add Rust parsers for tool calls.

matthewchan-g
authored and
copybara-github
committed

Commits on Jan 5, 2026

Internal code change.

ai-edge-bot
authored and
copybara-github
committed
Change huggingface_tokenizer_impl to a rust_library.

matthewchan-g
authored and
copybara-github
committed
Switch tool call parsing from ANTLR C++ to Rust.

matthewchan-g
authored and
copybara-github
committed

Commits on Jan 6, 2026

Commits on Jan 7, 2026

Add PromptTemplateCapabilities to GenericDataProcessor.

yichunk
authored and
copybara-github
committed
Add Reset functionality to AudioExecutor.

yichunk
authored and
copybara-github
committed

Commits on Jan 8, 2026

Fix a wrong scrub keyword

protobird-git
authored and
copybara-github
committed
Update dependencies of litert_lm

protobird-git
authored and
copybara-github
committed
Internal test update

ai-edge-bot
authored and
copybara-github
committed
Fix and refine benchmark timing.

hheydary
authored and
copybara-github
committed
Add support for using Engine::Session in litert_lm_lib.

hheydary
authored and
copybara-github
committed
Support running the model multiple times

tf-marissaw
authored and
copybara-github
committed
Do not print garbage during benchmarking

tf-marissaw
authored and
copybara-github
committed

Commits on Jan 9, 2026

Automated Code Change

ai-edge-bot
authored and
copybara-github
committed
Support configuring max output token.

ai-edge-bot
authored and
copybara-github
committed
Do not dump engine settings by default

tf-marissaw
authored and
copybara-github
committed
Update prebuilt binaries for gemma model constraint provider.

yishuangP
authored and
copybara-github
committed
Allow optional trailing commas in dict definitions in Python tool calls.

matthewchan-g
authored and
copybara-github
committed
Update the prompt input handling

ai-edge-bot
authored and
copybara-github
committed
Internal code change

ai-edge-bot
authored and
copybara-github
committed

Commits on Jan 10, 2026

Improve error message for LlmMetadata parsing failures.

ai-edge-bot
authored and
copybara-github
committed
[litertlm] fix prefill_preface_on_init

whhone
authored and
copybara-github
committed

Commits on Jan 12, 2026

Refactor step management in LlmLiteRtCompiledModelExecutor to use the actual current_step instead of referring from the processed tokens since processed tokens might be shared among different sessi…

ai-edge-bot
authored and
copybara-github
committed
Automated Code Change

Dave MacLachlan
authored and
copybara-github
committed
Sampler may handle next inference for better GPU performance

protobird-git
authored and
copybara-github
committed

Commits on Jan 13, 2026

[litertlm] Add a builder to scale ConversationConfig

whhone
authored and
copybara-github
committed
[litertlm] Remove JinjaPromptTemplate from SessionConfig

whhone
authored and
copybara-github
committed
[litertlm] Remove the non-default arguments from ConversationConfig::CreateDefault

whhone
authored and
copybara-github
committed
[litertlm] Add the C API to create Session with SessionConfig

whhone
authored and
copybara-github
committed
[litertlm] migrate ConversationConfig::CreateFromSessionConfig to builder (litert_lm)

whhone
authored and
copybara-github
committed

Commits on Jan 14, 2026

[litertlm] Make CreateFromSessionConfig private and rename to CreateInternal

whhone
authored and
copybara-github
committed
[litertlm][c] create conversation config with session config

whhone
authored and
copybara-github
committed
Avoid thread hopping with sync calls.

hheydary
authored and
copybara-github
committed
Internal changes ...

yishuangP
authored and
copybara-github
committed

Commits on Jan 15, 2026

Update dependencies of litert_lm

protobird-git
authored and
copybara-github
committed
[litertlm] consolidate the engine selection logic

whhone
authored and
copybara-github
committed
No public description

yichunk
authored and
copybara-github
committed
Initialize input tensor buffers to zero upon creation.

hheydary
authored and
copybara-github
committed

Commits on Jan 16, 2026

[litertlm-c]: Expose raw token counts and durations in BenchmarkInfo

ai-edge-bot
authored and
copybara-github
committed
[litertlm] add Kotlin API to set initial messages in Conversation

whhone
authored and
copybara-github
committed
[litertlm] update Kotlin README.md

whhone
authored and
copybara-github
committed

Commits on Jan 17, 2026

Internal change

ai-edge-bot
authored and
copybara-github
committed
Internal change

ai-edge-bot
authored and
copybara-github
committed
Refactor JSON tool parser to use Rust implementation.

matthewchan-g
authored and
copybara-github
committed
Internal changes and clean up.

ai-edge-bot
authored and
copybara-github
committed

Commits on Jan 18, 2026

[litertlm] refactor session config creation in C API

whhone
authored and
copybara-github
committed

Commits on Jan 20, 2026

Introduce EngineFactory for dynamic Engine creation.

hheydary
authored and
copybara-github
committed
Internal change and clean up.

ai-edge-bot
authored and
copybara-github
committed
internal change for chrome

ai-edge-bot
authored and
copybara-github
committed
Internal change

ai-edge-bot
authored and
copybara-github
committed
Decouple sampler data type from global precision.

fengwuyao
authored and
copybara-github
committed

Commits on Jan 21, 2026

Add GetTokens() to Tokenizer interface.

matthewchan-g
authored and
copybara-github
committed
Internal change and clean up.

ai-edge-bot
authored and
copybara-github
committed
Internal cleanup and changes.

ai-edge-bot
authored and
copybara-github
committed
Build fix on linux_arm64

protobird-git
authored and
copybara-github
committed
Update dependencies of litert_lm

protobird-git
authored and
copybara-github
committed
Reverts b781afee8f7f54cb249e922199d322906da1264b

ai-edge-bot
authored and
copybara-github
committed
[litertlm] add new tool type based on Open API in Kotlin

whhone
authored and
copybara-github
committed

Commits on Jan 22, 2026

Update dependencies of litert_lm

protobird-git
authored and
copybara-github
committed
Fix dangling reference issue.

hheydary
authored and
copybara-github
committed
Don't clear KV cache before prefill if requested.

protobird-git
authored and
copybara-github
committed
Add a flag to en/disable sampler to handle decode input tensors

protobird-git
authored and
copybara-github
committed

Commits on Jan 23, 2026

Internal change

whhone
authored and
copybara-github
committed
[litertlm] Add a Kotlin example of tool calling

whhone
authored and
copybara-github
committed
Internal change

ai-edge-bot
authored and
copybara-github
committed
Internal changes and clean up.

ai-edge-bot
authored and
copybara-github
committed
Disable gpu_madvise_original_shared_tensors when magic numbers are used.

protobird-git
authored and
copybara-github
committed
Change default backend to CPU in litert_lm_advanced_main.cc.

advaitjain
authored and
copybara-github
committed
Implement LlgConstraintProvider and ExternalConstraintProvider.

matthewchan-g
authored and
copybara-github
committed
Introduce LlmContext, configs and state definitions.

ai-edge-bot
authored and
copybara-github
committed

Commits on Jan 24, 2026

[litertlm] Update the instruction of setting hugging face token

whhone
authored and
copybara-github
committed
Encapsulate executor state within LlmContext.

ai-edge-bot
authored and
copybara-github
committed

Commits on Jan 25, 2026

Add force_f32 option in go & C api for better quality on gpu backend

protobird-git
authored and
copybara-github
committed
Update dependencies of litert_lm

protobird-git
authored and
copybara-github
committed

Commits on Jan 26, 2026

Commits on Jan 27, 2026

Add `support_single_turn` capability to prompt_template, which is enabled by checking whether there is corresponding `is_appending_to_prefill` in the jinja template. Templates with such capability …

yichunk
authored and
copybara-github
committed
Log the input.

ai-edge-bot
authored and
copybara-github
committed
Internal change

ai-edge-bot
authored and
copybara-github
committed
Use LFS (Large File Storage) to prebuilt binaries (#1294)
protobird-git
authored
Sync with github to support LFS (Large File Storage) for prebuilt binaries

protobird-git
authored and
copybara-github
committed
Revert "Sync with github to support LFS (Large File Storage) for prebuilt binaries" (#1300)
protobird-git
authored
Update dependencies with LFS

protobird-git
authored and
copybara-github
committed
[litertlm] Change ToolSet and OpenApiTool from abstract class to pure interface

whhone
authored and
copybara-github
committed
Internal change

ai-edge-bot
authored and
copybara-github
committed
Add OptionalArgs to Conversation SendMessage methods and enable multi prefill for Conversation API

yichunk
authored and
copybara-github
committed
Allow disabling caches from Android run script

tf-marissaw
authored and
copybara-github
committed

Commits on Jan 28, 2026

Internal change

ai-edge-bot
authored and
copybara-github
committed
Refactor Session Clone API to include synchronous and asynchronous versions.

ai-edge-bot
authored and
copybara-github
committed
Internal change

ai-edge-bot
authored and
copybara-github
committed
Simplify the build

advaitjain
authored and
copybara-github
committed