happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
llama.cpp/tools/server/README-dev.md

7.1 KiB
Raw Blame History

llama-server Development Documentation

This document provides an in-depth technical overview of llama-server, intended for maintainers and contributors.

If you are an end user consuming llama-server as a product, please refer to the main README instead.

Backend

Overview

The server supports two primary operating modes:

  • Inference mode: The default mode for performing inference with a single loaded GGUF model.
  • Router mode: Enables management of multiple inference server instances behind a single API endpoint. Requests are automatically routed to the appropriate backend instance based on the requested model.

The core architecture consists of the following components:

  • server_context: Holds the primary inference state, including the main llama_context and all active slots.
  • server_slot: An abstraction over a single “sequence” in llama.cpp, responsible for managing individual parallel inference requests.
  • server_routes: Middleware layer between server_context and the HTTP interface; handles JSON parsing/formatting and request routing logic.
  • server_http_context: Implements the HTTP server using cpp-httplib.
  • server_queue: Thread-safe queue used by HTTP workers to submit new tasks to server_context.
  • server_response: Thread-safe queue used by server_context to return results to HTTP workers.
  • server_response_reader: Higher-level wrapper around the two queues above for cleaner code.
  • server_task: Unit of work pushed into server_queue.
  • server_task_result: Unit of result pushed into server_response.
  • server_tokens: Unified representation of token sequences (supports both text and multimodal tokens); used by server_task and server_slot.
  • server_prompt_checkpoint: For recurrent (e.g., RWKV) and SWA models, stores snapshots of KV cache state. Enables reuse when subsequent requests share the same prompt prefix, saving redundant computation.
  • server_models: Standalone component for managing multiple backend instances (used in router mode). It is completely independent of server_context.
graph TD
    API_User <--> server_http_context
    server_http_context <-- router mode --> server_models
    server_http_context <-- inference mode --> server_routes
    server_routes -- server_task --> server_queue
    subgraph server_context
        server_queue --> server_slot
        server_slot -- server_task_result --> server_response
        server_slot[multiple server_slot]
    end
    server_response --> server_routes

TODO: mention about how batching is handled by server_slot

Thread Management

server_context runs on a dedicated single thread. Because it is single-threaded, heavy post-processing (especially after token generation) should be avoided, as it directly impacts multi-sequence throughput.

Each incoming HTTP request is handled by its own thread managed by the HTTP library. The following operations are performed in HTTP worker threads:

  • JSON request parsing
  • Chat template application
  • Tokenization
  • Conversion of server_task_result into final JSON response
  • Error formatting into JSON
  • Tracking of partial/incremental responses (e.g., streaming tool calls or reasoning steps)

Best practices to follow:

  • All JSON formatting and chat template logic must stay in the HTTP layer.
  • Avoid passing raw JSON between the HTTP layer and server_slot. Instead, parse everything into native C++ types as early as possible.

Testing

llama-server includes an automated test suite based on pytest.

The framework automatically starts a llama-server instance, sends requests, and validates responses.

For detailed instructions, see the test documentation.

Notable Related PRs

Web UI

The project includes a web-based user interface for interacting with llama-server. It supports both single-model (MODEL mode) and multi-model (ROUTER mode) operation.

The SvelteKit-based Web UI is introduced in this PR: https://github.com/ggml-org/llama.cpp/pull/14839

Features

  • Chat interface with streaming responses
  • Multi-model support (ROUTER mode) - switch between models, auto-load on selection
  • Modality validation - ensures selected model supports conversation's attachments (images, audio)
  • Conversation management - branching, regeneration, editing with history preservation
  • Attachment support - images, audio, PDFs (with vision/text fallback)
  • Configurable parameters - temperature, top_p, etc. synced with server defaults
  • Dark/light theme

Tech Stack

  • SvelteKit - frontend framework with Svelte 5 runes for reactive state
  • TailwindCSS + shadcn-svelte - styling and UI components
  • Vite - build tooling
  • IndexedDB (Dexie) - local storage for conversations
  • LocalStorage - user settings persistence

Architecture

The WebUI follows a layered architecture:

Routes → Components → Hooks → Stores → Services → Storage/API
  • Stores - reactive state management (chatStore, conversationsStore, modelsStore, serverStore, settingsStore)
  • Services - stateless API/database communication (ChatService, ModelsService, PropsService, DatabaseService)
  • Hooks - reusable logic (useModelChangeValidation, useProcessingState)

For detailed architecture diagrams, see tools/server/webui/docs/:

  • high-level-architecture.mmd - full architecture with all modules
  • high-level-architecture-simplified.mmd - simplified overview
  • data-flow-simplified-model-mode.mmd - data flow for single-model mode
  • data-flow-simplified-router-mode.mmd - data flow for multi-model mode
  • flows/*.mmd - detailed per-domain flows (chat, conversations, models, etc.)

Development

# make sure you have Node.js installed
cd tools/server/webui
npm i

# run dev server (with hot reload)
npm run dev

# run tests
npm run test

# build production bundle
npm run build

After public/index.html.gz has been generated, rebuild llama-server as described in the build section to include the updated UI.

Note: The Vite dev server automatically proxies API requests to http://localhost:8080. Make sure llama-server is running on that port during development.