Skip to content

Configuration Guide

Jump to bottom
sigoden edited this page Dec 10, 2024 · 10 revisions

Configuration File

When you run AIChat for the first time, it will automatically generate a default configuration file for you.

$ aichat
> No config file, create a new one? Yes
> Platform: openai
? API Key: ***
✨ Saved config file to '/home/alice/.config/aichat.yaml'

The location of config file is <user-config-dir>/aichat/config.yaml:

The config file path for each operating system are as follows:

  • Windows: C:\Users\Alice\AppData\Roaming\aichat\config.yaml
  • macOS: /Users/Alice/Library/Application Support/aichat/config.yaml
  • Linux: /home/alice/.config/aichat/config.yaml

You can also run the following command to get the config file path: aichat --info | grep config_file

You can modify the configuration file by referring to config.example.yaml.

LLM

These configurations mainly set the parameters of the LLM.

model: openai:gpt-4o             # Specify the LLM to use
temperature: null                # Set default temperature parameter
top_p: null                      # Set default top-p parameter, range (0, 1)

If we set the model to the client's name, such as model: openai, AIChat will choose the first chat model from the client's model list.

The range of temperature values is variable, mostly within [0, 1), but some are within [0, 2), such as Gemini and Qwen. Please pay attention to this when setting the values.

Behavior

These configurations mainly control some of aichat's behavior.

stream: true                     # Controls whether to use the stream-style API.
save: true                       # Indicates whether to persist the message
keybindings: emacs               # Choose keybinding style (emacs, vi)
editor: null                     # Specifies the command used to edit input buffer or session.yaml. env: EDITOR
wrap: no                         # Controls text wrapping (no, auto, <max-width>)
wrap_code: false                 # Enables or disables wrapping of code blocks

Function calling

These configurations mainly for function calling.

# Visit https://github.com/sigoden/llm-functions for setup instructions
function_calling: true           # Enables or disables function calling (Globally).
mapping_tools:                   # Alias for a tool or toolset
  fs: 'fs_cat,fs_ls,fs_mkdir,fs_rm,fs_write'
use_tools: null                  # Which tools to use by default

If we set function_calling: false, it globally disables all function calling.

The mapping_tools configures the grouping of tools for ease of use tools later. For example, we can create an fs group that includes all fs-related tools.

If we set use_tools: fs, it's easy to use all fs-related tools.

Prelude

AIChat also have many working states: normal, role, session, rag, and agent.

The prelude controls which state AIChat starts in.

The following are configuration items related to prelude:

prelude: null                    # Set a default role or session to start with (e.g. role:<name>, session:<name>, <session>:<role>)
repl_prelude: null               # Overrides the `prelude` setting specifically for conversations started in REPL
agent_prelude: null              # Set a session to use when starting a agent. (e.g. temp, default)

If we want to start in a coder role, we can set prelude: "role:coder".

If we want to start in a coding session, we can set prelude: "session:coding".

If we want to start in a %shell% role only when entering the REPL, we can use repl_prelude: "role:%shell%".

The agent prelude only takes effect when entering the agent. If we set agent_prelude: default, it means that the session named "default" will be automatically loaded when entering the agent.

Session

These configurations mainly set the parameters of the session.

# Controls the persistence of the session. if true, auto save; if false, not save; if null, asking the user
save_session: null
# Compress session when token count reaches or exceeds this threshold
compress_threshold: 4000
# Text prompt used for creating a concise summary of session message
summarize_prompt: '...'
# Text prompt used for including the summary of the entire session
summary_prompt: '...'

If leave save_session as null, AIChat will prompt you to save the session when you exit.

If save_session: true is set, AIChat will automatically save the session upon exit.

If save_session: false is set, AIChat will exit without saving the session.

When the chat history tokens exceed compress_threshold, AIChat automatically compresses the session. summarize_prompt is used to guide the LLM in summarizing the chat history, and summary_prompt is used to include the history summary.

RAG

These configurations mainly set the parameters of the RAG.

rag_embedding_model: null                   # Specifies the embedding model to use
rag_reranker_model: null                    # Specifies the rerank model to use
rag_top_k: 4                                # Specifies the number of documents to retrieve
rag_chunk_size: null                        # Specifies the chunk size
rag_chunk_overlap: null                     # Specifies the chunk overlap
rag_min_score_vector_search: 0              # Specifies the minimum relevance score for vector-based searching
rag_min_score_keyword_search: 0             # Specifies the minimum relevance score for keyword-based searching
# Defines the query structure using variables like __CONTEXT__ and __INPUT__ to tailor searches to specific needs
rag_template: |
  ...
# Define document loaders to control how RAG and `.file`/`--file` load files of specific formats.
document_loaders:
  ...

See RAG-Guide for more details.

Apperence

These configurations mainly affect the appearance.

highlight: true                  # Controls syntax highlighting
light_theme: false               # Activates a light color theme when true. env: AICHAT_LIGHT_THEME
# Custom REPL prompt, see https://github.com/sigoden/aichat/wiki/Custom-REPL-Prompt for more details
left_prompt: '...'
right_prompt: '...'

Misc

serve_addr: 127.0.0.1:8000                  # Default serve listening address 
user_agent: null                            # Set User-Agent HTTP header, use `auto` for aichat/<current-version>
save_shell_history: true                    # Whether to save shell execution command to the history file

Agent-specific

The following are the unique configuration items for the agent:

model: openai:gpt-4o             # Specify the LLM to use
temperature: null                # Set default temperature parameter
top_p: null                      # Set default top-p parameter, range (0, 1)
use_tools: null                  # Which additional tools to use by agent. (e.g. 'fs,web_search')
agent_prelude: null              # Set a session to use when starting the agent. (e.g. temp, default)

variables:                       # Custom default values for the agent variables
  <key>: <value>

The agent configuration file is located at <aichat-config-dir>/agents/<agent-name>/config.yaml,

You can obtain the agent configuration directory by using the command aichat --agent <agent-name> --info | grep data_dir.

Clients

The config.example.yaml provides detailed configuration examples for various API providers. You can copy and paste these examples directly into your configuration.

Add chat models

clients:
  - type: <client-type>
    ...
    models:
      - name: xxxx
        max_input_tokens: 100000
        supports_vision: true
        supports_function_calling: true

Add embedding models

  - type: <client-type>
    ...
    models:
      - name: xxxx
        type: embedding
        max_input_tokens: 200000
        max_tokens_per_chunk: 2000
        default_chunk_size: 1500                        
        max_batch_size: 100

type: embedding indicates an embedding model.

Add reranker models

  - type: <client-type>
    ...
    models:
      - name: xxxx
        type: reranker 
        max_input_tokens: 2048

type: reranker indicates a reranking model.

Set socks/http(s) proxy

clients:
  - type: <client-type>
    ...
    extra:
      proxy: socks5://127.0.0.1:1080 # or http://127.0.0.1:8080

Patch API request

AIChat supports patching request url, headers and body.

clients:
  - type: <client-type>
    ...
    patch:                                          # Patch api
      chat_completions:                             # Api type, possible values: chat_completions, embeddings, and rerank
        <regex>:                                    # The regex to match model names, e.g. '.*' 'gpt-4o' 'gpt-4o|gpt-4-.*'
          url: ''                                   # Patch request url
          body:                                     # Patch request body
            <json>
          headers:                                  # Patch request headers
            <key>: <value>

Patch gemini for using the most relaxed security settings

clients:
  - type: gemini
    ...
    patch:
      chat_completions:
        '.*':
          body:
            safetySettings:
              - category: HARM_CATEGORY_HARASSMENT
                threshold: BLOCK_NONE
              - category: HARM_CATEGORY_HATE_SPEECH
                threshold: BLOCK_NONE
              - category: HARM_CATEGORY_SEXUALLY_EXPLICIT
                threshold: BLOCK_NONE
              - category: HARM_CATEGORY_DANGEROUS_CONTENT
                threshold: BLOCK_NONE

Patch deepseek for using the beta API usage

```yaml
clients:
  - type: openai-compatible
    name: deepseek
    ...
    patch:
      chat_completions:
        '.*':
          url: https://api.deepseek.com/beta/chat/completions

Add client for openai-compatible API provider

clients:
  - type: openai-compatible
    name: local
    api_base: http://localhost:8080/v1
    api_key: xxx                                      # Optional
    models:
      - name: llama3.1
        max_input_tokens: 128000
        supports_function_calling: true
      - name: jina-embeddings-v2-base-en
        type: embedding
        default_chunk_size: 1500
        max_batch_size: 100
      - name: jina-reranker-v2-base-multilingual
        type: reranker
Clone this wiki locally