darknoah/aime-chat/aime-chat-docs

Aime Chat Docs

Use this skill when answering questions about Aime Chat configuration and feature conventions.

Reference Index

• Secret/key management: Read references/secrets.md when managing individual records in Settings → Secrets: list/filter, inspect, create, rename, replace values, edit descriptions, toggle global environment injection or delete. Use metadata for routine management and explicit per-record reveal only when the task requires the value.
• Application doctor: Read references/doctor.md when diagnosing application status, API port/health, network proxy, local model downloads, default models, provider enablement/API bases/active models, or runtime availability. Run the read-only doctor first and use the relevant configuration reference for any follow-up repair.
• Assistant personalities: Read references/create-personality.md when explaining how to create or edit an assistant personality, how SOUL.md front matter works, or where avatars live.
• Local voices: Read references/create-local-voice.md when explaining how to create local voice folders under user data, what files are required, or how voice IDs are derived.
• Skill management: Read references/manage-skills.md when explaining how to create or add Aime Chat skills globally or for the current project.
• Runtime environments: Read references/runtime.md when listing runtime status, paths and versions, or installing, reinstalling or uninstalling UV/Python, Bun, Node.js, PaddleOCR, QwenAudio or Agent Browser.
• Available models: Read references/get-available-models.md when explaining how to list configured provider models through the local API server.
• Default models: Read references/default-models.md when listing, setting or clearing the global default chat, fast, vision, embedding, reranker, OCR, transcription, speech, image or video model.
• Local models and knowledge bases: Read references/local-models.md when listing/downloading/deleting local models or preparing a knowledge base. Default downloads to ModelScope unless the user chooses another source or the catalog does not offer it. Do not set a download timeout; omit --timeout unless the user explicitly requests one. If local embedding or reranker models are missing, guide their download, verify completion, then create the knowledge base with the selected provider model IDs.
• Available Agents: Read references/get-available-agents.md when explaining how to list enabled Agents through the local API server.
• Agent detail: Read references/get-agent.md when explaining how to fetch a single Agent's full config (instructions, tools, sub-agents, suggestions) by id through the local API server.
• Create/update Agent: Read references/save-agent.md when explaining how to create a new Agent or update an existing one (name, instructions, tools, model, tags) through the local API server.
• Available tools: Read references/get-available-tools.md when explaining how to list enabled tools through the local API server.
• Tool detail: Read references/get-tool.md when explaining how to fetch a single tool's detail (sub-tools, input schema, status) by id through the local API server.
• PTC (Programmatic Tool Calling): Read references/use-ptc.md when explaining how to use the CodeExecution PTC mode to call tools programmatically in code, batch tool calls in a loop, call ChatCompletion inside code, or report progress via the Message tool.
• Create thread: Read references/create-thread.md when explaining how to create a new chat thread (optionally under a project, with an Agent and model) through the local API server.
• Send thread message: Read references/send-thread-message.md when sending text and/or local images to an existing idle thread and waiting for the final assistant text through the local API server.
• Running threads: Read references/list-running-threads.md when explaining how to list threads that are currently streaming (running) through the local API server.
• Projects: Read references/list-projects.md when explaining how to list projects (with pagination and title filter) through the local API server.
• Chat history: Read references/chat-history.md when explaining how to list recent chat threads, read a thread's messages, or keyword-search across chat history through the local API server.
• Evaluation datasets: Read references/manage-eval-datasets.md when creating datasets, adding test cases, or importing/exporting CSV and JSONL samples.
• Evaluation scorers: Read references/create-eval-scorers.md when creating, testing, or updating deterministic checks and LLM judges.
• Evaluation runs: Read references/run-eval-experiments.md when running dataset tests, waiting for results, comparing experiments, or scoring existing conversations. For a full evaluation workflow, use all three evaluation references.

API Scripts

The scripts/ folder contains standalone Python scripts (standard library only, Python 3) for the Aime Chat local API. Scripts read the API base URL from the AIME_CHAT_API_BASE_URL environment variable and require the local API server to be enabled and running. The doctor can also diagnose an unavailable server using that environment-provided address. Do not assume or hardcode an API port. Prefer running these scripts over hand-writing HTTP requests.

• [scripts/secrets.py](scripts/secrets.py): Manage individual secrets using list [--key NAME] [--global true|false], get --id ID [--reveal], create --file FILE, update --id ID --file FILE and delete --id ID. Use --file - for JSON on stdin; normal responses omit secret values. Supports --timeout on each command.
• [scripts/doctor.py](scripts/doctor.py): Check API TCP/health, applied app proxy and shell proxy settings, local downloads, default models, every configured provider, and runtimes. Reads the API address exclusively from AIME_CHAT_API_BASE_URL. Human-readable output by default; use --json for a complete report or --cached for cached runtime state. No configuration changes or model inference.
• [scripts/runtime.py](scripts/runtime.py): List runtime environments with list (or list --cached) and manage a runtime with install <pkg>, reinstall <pkg> or uninstall <pkg>. Returns JSON; supports --timeout in seconds. Check supportedActions and dependencies before mutations.
• [scripts/default_models.py](scripts/default_models.py): List effective defaults with list, or update selected fields with set --model <id> --fast-model <id> and the other model options shown by set --help. Pass an empty string to clear a field; unspecified fields are preserved. Returns JSON.
• [scripts/local_models.py](scripts/local_models.py): List local model status with list --type embedding or list --type reranker; download with download --type <type> --model-id <id> (defaults to ModelScope; use --source huggingface to override), or delete with delete --type <type> --model-id <id>. All commands default to no HTTP timeout. Downloads wait for completion; recheck status before creating a knowledge base.
• [scripts/evals.py](scripts/evals.py): Manage evaluation datasets, samples, scorers and experiments. Use --help or <command> --help; writes JSON to stdout. Body commands take --file request.json (or --file - for stdin). start-experiment accepts --workspace and fills a missing JSON workspace from the dataset's server-provided default. start-experiment and get-experiment support --wait, --wait-timeout and --poll-interval.
• [scripts/get_available_models.py](scripts/get_available_models.py): List configured provider models. Supports --type (llm, embedding, reranker, image_generation, video_generation, transcription, speech, ocr, music) and --json for raw output.
• [scripts/get_available_agents.py](scripts/get_available_agents.py): List enabled Agents as - [<agent-id>]: <description>. Supports --visible-only to exclude hidden Agents and --json for raw output.
• [scripts/get_agent.py](scripts/get_agent.py): Get a single Agent's full config by id, e.g. python scripts/get_agent.py --id code-agent. Shows type, description, tools, sub-agents, suggestions, and instructions. Supports --json for raw output.
• [scripts/save_agent.py](scripts/save_agent.py): Create or update an Agent (always saved as active). Requires --id (letters/digits/-/_ only), --name, --description, --instructions. Optional repeatable --suggestion, --tool, and --sub-agent.
• [scripts/get_available_tools.py](scripts/get_available_tools.py): List enabled tools grouped by type (mcp, build-in, skill), expanding toolkit sub-tools. Descriptions are truncated to the first 100 characters. Supports --json for raw output.
• [scripts/get_tool.py](scripts/get_tool.py): Get a single tool's detail by id, e.g. python scripts/get_tool.py --id skill:local:xlsx. Shows type, status, description, and sub-tools. Supports --json for raw output.
• [scripts/preview_git_skill.py](scripts/preview_git_skill.py): Preview the skills available in a git repository, e.g. python scripts/preview_git_skill.py https://github.com/<owner>/<repo>.
• [scripts/import_skills.py](scripts/import_skills.py): Import skills globally or into the current project. Use --repo-or-url with optional repeatable --skill for repo installs, --file for .skill/.zip packages, and --path <cwd> for project-scoped installs (omit --path for global).
• [scripts/create_thread.py](scripts/create_thread.py): Create a new chat thread. Optional --project-id, --agent-id, --model, and --json for raw output.
• [scripts/send_thread_message.py](scripts/send_thread_message.py): Send text and/or repeatable local --image files to an existing idle thread through the existing chat() API. Waits for chat() to finish, then prints text from the last returned message. Throws an error instead of queueing when the thread is busy.
• [scripts/list_running_threads.py](scripts/list_running_threads.py): List currently running (streaming) threads. Supports --json for raw output.
• [scripts/list_projects.py](scripts/list_projects.py): List projects. Optional --filter, --page, --size, and --json for raw output.
• [scripts/chat_history_list.py](scripts/chat_history_list.py): List recent chat threads (normal and project threads grouped by project). Optional --since, --until, --limit, and --include-cron (cron-created threads are excluded by default).
• [scripts/chat_history_read.py](scripts/chat_history_read.py): Read messages of a single thread by --thread-id (thread meta + plain-text messages). Optional --limit, --since for delta reads, and --include-tools for tool-call summaries.
• [scripts/chat_history_search.py](scripts/chat_history_search.py): Keyword search across recent chat threads with --query (space-separated keywords are fuzzy-matched, all must appear), matching message content, thread titles and project names. Excerpts are grouped per thread with project info (if any), thread id and title. Optional --since, --limit, and --thread-limit.