Chat
chat.ionclocklab.comThe main AI web app. Use it for chat, coding, research, writing, model comparison, image generation, image editing, web search, speech-to-text, and text-to-speech.
Open WebUILoginModelsImagesAudio
Ion Clock Lab Platform
Ion Clock Lab Service Guide
A unified guide for Ion Clock Lab services: Chat, Notes, LLM Gateway, API Docs, model families, image generation, image editing, Brave Search, Whisper transcription, and TTS.
chat.ionclocklab.comnotes.ionclocklab.comllm.ionclocklab.comdocs.ionclocklab.com
1. Main Sites
Notes
notes.ionclocklab.comA private Trilium Notes knowledge base for lab notes, experiment logs, meeting records, project planning, and structured long-term knowledge.
TriliumKnowledge BaseResearch Notes
LLM Gateway
llm.ionclocklab.comThe LiteLLM API gateway. Developers can call different model providers through one OpenAI-compatible API base URL.
LiteLLMOpenAI APIVirtual Keys
API Docs
docs.ionclocklab.comTechnical API documentation for chat completions, images, embeddings, speech, transcription, and public model lists.
DocsEnglish中文Models
2. Chat Features
Accounts and login
Users access the platform through individual accounts. The admin can control signup, user permissions, model visibility, and feature access.
Model picker
Users can choose among GPT-style models, Claude, Gemini, Grok, DeepSeek, Qwen, GLM, Kimi, MiniMax, and other model families exposed through LiteLLM.
Image generation
The platform supports image creation through gpt-image-2. Users can ask for diagrams, logos, visual concepts, and scientific illustrations.
Image editing
Image editing is enabled through the OpenAI-compatible /v1/images/edits endpoint. It has been tested successfully through LiteLLM and gpt-image-2.
Web search
Brave Search / brave_llm_context is enabled for current information and source-based answers. Use it when asking about recent news, current documentation, or online references.
Speech-to-text
Whisper and gpt-4o-mini-transcribe can convert microphone or uploaded audio into text, useful for quick dictation and voice input.
Text-to-speech
tts-1 can read assistant responses aloud. Response splitting by punctuation keeps long replies natural and stable.
Research workflow
Use Chat for active thinking and problem solving, then move stable long-term notes into Trilium Notes.
3. Model Families
GPT / OpenAI-style
Good default for daily work: chat, coding, summarization, structured writing, quick reasoning, and tool use. Typical models include gpt-4o-mini and gpt-4.1-mini.
Claude
Strong for careful writing, long-form analysis, code explanation, document reasoning, and nuanced answers. Claude Sonnet, Haiku, and Opus variants may appear in the picker.
Gemini
Useful for broad multimodal reasoning, long-context tasks, research summaries, and general assistant work.
Grok
Useful for conversational tasks, brainstorming, current-events style prompts, and alternative reasoning perspectives.
DeepSeek
Useful for technical reasoning, programming, math-like tasks, and cost-effective problem solving.
Qwen
Strong for Chinese-English tasks, multilingual writing, coding, and general-purpose assistant workflows.
GLM
Good bilingual Chinese-English general model family for writing, summarization, translation-like tasks, and technical chat.
Kimi
Useful for long-context reading, Chinese and English document understanding, research notes, and summarization.
MiniMax
Alternative general model family for creative writing, chat, and flexible assistant tasks.
Image / Audio
gpt-image-2 handles image generation and editing. whisper-1 and gpt-4o-mini-transcribe handle transcription. tts-1 handles speech output.