MythEclipse/asepharyana-hub-llm-api
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
17 Commits 2 Branches 0 Tags
5cba76d280079964d52733f888c246b37cbe4b7b
Commit Graph
17 Commits
This Branch
This Branch
All Branches
Author SHA1 Message Date
Asep Haryana 5cba76d280 fix(stream): separate reasoning_content from content in SSE streaming 
- Before </think>: tokens sent as delta.reasoning_content
- After </think>: tokens sent as delta.content
- Handles </think> split across token boundaries
- Clean <|im_start|>/<|im_end|>/<think>/</think> from streamed text
- Normal models (no thinking) send all as content as before
 2026-07-26 19:04:29 +07:00
Asep Haryana 6ff31b5f62 feat: render GGUF Jinja template via minijinja crate 
- Replaced manual prompt building with GGUF's chat template rendered through minijinja (Rust Jinja2 engine)
- Simplified template: removed multi-step tool detection, reasoning extraction (not needed at template level)
- Added reasoning_content separation: clean_text() returns (reasoning, answer) tuple
- Added reasoning_content field to ResponseMessage and SseDelta for OpenAI-compatible output
- Embedded template at build time via include_str! from templates/chat_template.jinja
- Built-in support for enable_thinking, tool_definitions, tool_calls, tool_response
 2026-07-26 18:21:35 +07:00
Asep Haryana 9a63ff1601 feat: use GGUF chat template via apply_chat_template 
- Replaced manual prompt building with LlamaModel::apply_chat_template
- Uses model's baked-in Jinja template (system/user/assistant/tool format)
- Added <think> trigger after template for thinking mode
- Clean_text strips only <|im_end|>, <|im_start|>, <think>, </think>
 2026-07-26 17:10:42 +07:00
Asep Haryana 59c77108a5 fix: skip leading EOS tokens in streaming + non-streaming 
- Skip <|im_end|> generated as first token (prevents empty responses)
- Clean <think> tags as plain text in generated output
 2026-07-26 16:09:04 +07:00
Asep Haryana 495b9ed126 fix: remove /think trigger, add back <think> in clean_text 
- Removed /think trigger from prompt (model generates better without it)
- Added <think> and </think> back to clean_text (model uses plain text
  thinking tags, not special tokens)
 2026-07-26 16:03:18 +07:00
Asep Haryana 254532458b Revert "fix: remove /think trigger from prompt, use plain assistant prefix" 
This reverts commit 7c8f747faf.
 2026-07-26 15:49:04 +07:00
Asep Haryana 7c8f747faf fix: remove /think trigger from prompt, use plain assistant prefix 2026-07-26 15:48:44 +07:00
Asep Haryana d8425ea3b3 feat: switch to MiniCPM5-1B-Claude-Opus-Fable5-V2-Thinking-Q8_0 GGUF 
- Updated model path + model ID for new 1B thinking model
- Updated prompt builder to use MiniCPM5 native /think trigger
- Updated clean_text to strip MiniCPM5 special tokens
- Bumped n_ctx from 2048 to 8192
 2026-07-26 15:38:05 +07:00
asepharyana 14032db705 ci: add workflow_dispatch to notify-parent for manual triggering 2026-07-25 18:13:35 +07:00
asepharyanaandClaude Code 2a5ab8b6d5 ci(llm-api): add notify-parent workflow 
Trigger repository_dispatch to parent hub repo on push to master.

Co-Authored-By: Claude Code <noreply@anthropic.com>
 2026-07-25 15:11:38 +07:00
asepharyanaandGitHub 7e717cc808 Merge pull request #1 from asepharyana/refactor/clean-architecture 
refactor(llm-api): implement clean architecture following scraper pat…
 2026-07-25 15:09:04 +07:00
asepharyanaandClaude Code e351d74fa4 refactor(llm-api): implement clean architecture following scraper pattern 
Split monolithic 1012-line main.rs into layered hexagonal architecture:
- Domain: entity types and LlmError enum
- Application: prompt building, sampler construction, tool call parsing
- Infrastructure: LlamaEngine wrapping llama-cpp-2 with isolated unsafe transmute
- Presentation: Axum handlers, middleware (auth), error chain, router
- Config: type-safe AppConfig with LazyLock
- Bootstrap: Application struct with build() + run()

Resolves build_sampler/build_sampler_params duplication.
Adds simple web chat UI at GET /.

Co-Authored-By: Claude Code <noreply@anthropic.com>
 2026-07-25 15:07:19 +07:00
asepharyana dfd6fa66a7 chore: remove infra files managed by hub monorepo 
Dockerfile, docker-compose.yml, llm-api.yml, and traefik-router.yaml
are now managed centrally in asepharyana-hub's infra/ directory.
 2026-07-25 12:58:12 +07:00
Asep Haryana c77ceee83f feat: OpenAI-compatible tool calling + sampling params 
- Tool calling: tools array, tool_choice, tool call parsing, round-trip
- Sampling: temperature, top_p, top_k, min_p, seed
- Penalties: repeat_penalty, frequency_penalty, presence_penalty
- Stop sequences support
- Chat history with assistant tool_calls and tool responses
- GenerateStream with SendSampler wrapper for thread-safe async
 2026-07-25 11:42:57 +07:00
Asep Haryana e1f5195407 feat: add SSE streaming support (OpenAI-compatible) 
- Token-by-token streaming via Server-Sent Events
- tokio::sync::mpsc channel + ReceiverStream for clean async
- OpenAI SSE format: role chunk → content chunks → finish chunk
- Non-streaming still works with the same code path
- Uses spawn_blocking pattern for CPU-bound inference
 2026-07-25 11:33:59 +07:00
Asep Haryana 6edf6cee5d fix: use token_to_piece_bytes for proper token decoding 2026-07-25 11:17:58 +07:00
Asep Haryana 33b42b789c feat: initial release - OpenAI-compatible LLM server with llama-cpp-2 2026-07-25 11:00:03 +07:00