# LiteRT-LM | google-ai-edge | Knowledge Base | Instagit

GitHub Stars: 1.7k

Repository: https://github.com/google-ai-edge/LiteRT-LM

---

## Articles

### [LiteRT-LM Engine, Session, and Conversation Objects: Key Differences Explained](/google-ai-edge/LiteRT-LM/lrt-lm-differences-engine-session-conversation-objects)

Understand the key differences between LiteRT-LM Engine, Session, and Conversation objects. Learn how Engine creates Sessions for low-level control and Conversations for high-level chat abstraction and multi-turn history manage...

- Tags: deep-dive
- Published: 2026-04-06

### [Supported Model Formats for LiteRT-LM: Conversion Requirements Explained](/google-ai-edge/LiteRT-LM/lrt-lm-supported-model-formats-conversion-requirements)

Explore LiteRT-LM supported model formats including tflite, task, and litertlm. Understand conversion requirements for efficient on-device deployment. Turn your models into production-ready litertlm files.

- Tags: architecture
- Published: 2026-04-06

### [Thread Model for Handling Asynchronous Inference Operations in LiteRT-LM](/google-ai-edge/LiteRT-LM/lrt-lm-thread-model-async-inference)

Explore the LiteRT-LM thread model for asynchronous inference. Discover how its task-centric pipeline with dedicated ThreadPools ensures non-blocking LLM operations and efficient callback dispatch.

- Tags: internals
- Published: 2026-04-06

### [How to Build LiteRT-LM from Source Using Bazel or CMake](/google-ai-edge/LiteRT-LM/lrt-lm-build-from-source-bazel-cmake)

Learn to build LiteRT-LM from source with Bazel or CMake. Compile the inference engine easily for your AI projects. Get started today!

- Tags: how-to-guide
- Published: 2026-04-06

### [How LiteRT-LM Manages Multi-Turn Conversation State: Architecture and Implementation](/google-ai-edge/LiteRT-LM/lrt-lm-multi-turn-conversation-state-management)

Learn how LiteRT-LM manages multi turn conversation state using its C++ Conversation class for efficient chat history and KV cache management. Discover the architecture.

- Tags: architecture
- Published: 2026-04-06

### [How to Use Streaming Inference with LiteRT-LM's Callback-Based API](/google-ai-edge/LiteRT-LM/lrt-lm-streaming-inference-callback-api)

Learn how LiteRT-LM's callback-based API enables non-blocking streaming inference for real-time LLM output processing.

- Tags: how-to-guide
- Published: 2026-04-06

### [LiteRT-LM Sampler Parameters: Complete Guide to TopK, TopP, Greedy, and Temperature](/google-ai-edge/LiteRT-LM/lrt-lm-sampler-parameters-topk-topp-greedy-temperature)

Master LiteRT-LM sampler parameters like TopK, TopP, Greedy, and Temperature. Optimize your text generation with this complete guide to LiteRT-LM's SamplerParameters.

- Tags: api-reference
- Published: 2026-04-06

### [How to Configure Backend Options for LiteRT-LM: CPU, GPU, Vision, and Audio Backends](/google-ai-edge/LiteRT-LM/lrt-lm-configure-backend-options-cpu-gpu-vision-audio)

Learn how to configure LiteRT-LM backend options for CPU, GPU, vision, and audio. Control inference engines via CLI or Python API for optimized performance.

- Tags: how-to-guide
- Published: 2026-04-06

### [LiteRT-LM Multimodal Input Types: Text, Images, and Audio Support](/google-ai-edge/LiteRT-LM/lrt-lm-supported-multimodal-input-types-text-image-audio)

Explore LiteRT-LM's multimodal input capabilities. Learn how it processes text, images, and audio for advanced AI applications.

- Tags: getting-started
- Published: 2026-04-06

### [How the C API Manages the Lifecycle of LiteRT-LM Engine and Sessions](/google-ai-edge/LiteRT-LM/lrt-lm-c-api-engine-session-lifecycle-management)

Learn how the LiteRT-LM C API manages engine and session lifecycles with explicit create/delete functions for clear ownership from start to finish.

- Tags: internals
- Published: 2026-04-06

### [LoRA Adapter Architecture in LiteRT-LM: Runtime Integration Deep Dive](/google-ai-edge/LiteRT-LM/lrt-lm-lora-adapter-integration-architecture)

Explore the six-layer LoRA adapter architecture in LiteRT-LM. Learn how LiteRT-LM integrates LoRA for efficient runtime adaptation via memory-mapped files, regex validation, and manager components.

- Tags: deep-dive
- Published: 2026-04-06

### [How LiteRT-LM Handles Task Cancellation and Sequential Execution in the ExecutionQueue](/google-ai-edge/LiteRT-LM/lrt-lm-executionqueue-task-cancellation-sequential-execution)

Discover how LiteRT-LM ensures task cancellation and sequential execution with its FIFO ExecutionQueue. Learn about monotonic IDs and the safe Remove API.

- Tags: internals
- Published: 2026-04-06