codecompanion-history.nvim

repository·main·Indexed 41 days ago

https://github.com/ravitemer/codecompanion-history.nvim

A history management extension for codecompanion.nvim that allows users to save, browse, and restore chat sessions. It preserves full context including tools, variables, LLM settings, and pinned references. Features include an automatic title generation system, a summary system, and a vector-based memory system for semantic search across past conversations via the @memory tool (requires VectorCode CLI). Supports integration with telescope.nvim, snacks.nvim, and fzf-lua for enhanced picker interfaces.

Tokens: 3.3K
Snippets: 6
Records: 10
Agent score: 2%

What's inside codecompanion-history.nvim

How Chat Restoration Works
main
When you restore a chat session from history, the extension recreates the full context. The following features are preserved:
- Messages History: All previous messages.
- System Prompts: The original system prompt used.
- LLM Settings: The specific adapter, model, temperature, and other settings.
- Tools: Tool schemas, system prompts, and execution results (outputs).
- Variables & References: Variables used and code snippets/command outputs added via slash commands.
- Images: Restored as base64 strings.
- Pinned References: Any references pinned during the session.
Note on Watchers: While watched buffer states are saved, they require the original buffer context to resume watching functionality.
https://github.com/ravitemer/codecompanion-history.nvim/blob/main/README.md
How the Memory System works with @memory
main
The Memory System provides a vector-based search capability for your chat history. It works by using the VectorCode CLI to index chat summaries.
1. Summarization: When a chat summary is generated (via gcs), it can be automatically indexed if auto_create_memories_on_summary_generation is enabled.
2. Indexing: Summaries are processed and stored in a vector database managed by VectorCode.
3. Retrieval: Once indexed, the @memory tool becomes available in new chats. You can use it to perform semantic searches across your past conversations to bring relevant context into your current session.
This requires the vectorcode executable to be available on your system path.
https://github.com/ravitemer/codecompanion-history.nvim/blob/main/README.md
Manage Chat History and Summaries
main
Commands
- :CodeCompanionHistory: Open the history browser.
- :CodeCompanionSummaries: Browse all generated summaries.
Chat Buffer Keymaps (Default)
- gh: Open history browser.
- sc: Save current chat manually.
- gcs: Generate summary for current chat.
- gbs: Browse saved summaries.
History Browser Actions
In the history browser, use <CR> to open a chat. Use the following keys to manage entries:
Action Normal Mode Insert Mode
Delete d <M-d>
Rename r <M-r>
Duplicate <C-y> <C-y>
Summary Browser Actions
In the summary browser, use <CR> to add the selected summary to the current chat. Use the following keys to manage entries:
Action Normal Mode Insert Mode
Delete d <M-d>
https://github.com/ravitemer/codecompanion-history.nvim/blob/main/README.md
Install CodeCompanion History Extension
main
To use this extension, you must first install it as a dependency of codecompanion.nvim. Using lazy.nvim, add ravitemer/codecompanion-history.nvim to your codecompanion.nvim dependencies.
Requirements
- Neovim >= 0.8.0
- codecompanion.nvim
- VectorCode CLI (Optional: required for the @memory tool)
- snacks.nvim, telescope.nvim, or fzf-lua (Optional: for enhanced picker interfaces)
```
{
    "olimorris/codecompanion.nvim",
    dependencies = {
        --other plugins
        "ravitemer/codecompanion-history.nvim"
    }
}
```
https://github.com/ravitemer/codecompanion-history.nvim/blob/main/README.md
Configure the Title Refresh feature
main
The extension can automatically update chat titles as the conversation progresses. You can control this via title_generation_opts.
- refresh_every_n_prompts: Frequency of refresh (e.g., every 3rd user prompt).
- max_refreshes: Maximum number of times a title can be updated to prevent excessive API usage.
Note: Individual messages are truncated at 1000 characters, and total conversation context is limited to 10,000 characters for title generation.
```
title_generation_opts = {
    refresh_every_n_prompts = 3, -- Refresh after every 3rd user prompt
    max_refreshes = 10,           -- Allow up to 10 refreshes per chat
}
```
https://github.com/ravitemer/codecompanion-history.nvim/blob/main/README.md

Action	Normal Mode	Insert Mode
Delete	`d`	`<M-d>`
Rename	`r`	`<M-r>`
Duplicate	`<C-y>`	`<C-y>`

Action	Normal Mode	Insert Mode
Delete	`d`	`<M-d>`

Configure the @memory tool with VectorCode

main

If you have the VectorCode CLI installed, this plugin creates a @memory tool. This tool allows the LLM to search through previous chat summaries to provide context in new conversations.

You can configure the memory submodule using the opts.memory table in your setup.

opts.memory = {
    auto_create_memories_on_summary_generation = true,
    -- path to the `vectorcode` executable
    vectorcode_exe = "vectorcode",
    tool_opts = {
        -- default number of memories to retrieve
        default_num = 10 
    },
    -- whether to enable notification
    notify = true,
    -- whether to automatically update the index of all existing memories on startup
    -- (requires VectorCode 0.6.12+ for efficient incremental indexing)
    index_on_startup = false,
}

Configure CodeCompanion History Extension

main

The extension is configured within the codecompanion.nvim setup function under the extensions.history key.

Key Configuration Areas

Chat Management

auto_save: Boolean to enable/disable automatic session saving.
keymap: Keymap to open history from chat buffer (default: gh).
save_chat_keymap: Keymap to save chat manually (default: sc).
picker: Choose between "telescope", "snacks", "fzf-lua", or "default".
auto_generate_title: Enables automatic title generation.
title_generation_opts: Configure adapter, model, and refresh_every_n_prompts for titles.

Summary System

create_summary_keymap: Keymap to generate summary (default: gcs).
browse_summaries_keymap: Keymap to browse summaries (default: gbs).
generation_opts: Configure adapter, model, context_size, and system_prompt for summaries.

Memory System (Requires VectorCode CLI)

auto_create_memories_on_summary_generation: Automatically index summaries.
vectorcode_exe: Path to the vectorcode executable.
index_on_startup: Whether to index existing memories on startup.

require("codecompanion").setup({
    extensions = {
        history = {
            enabled = true,
            opts = {
                keymap = "gh",
                save_chat_keymap = "sc",
                auto_save = true,
                expiration_days = 0,
                picker = "telescope",
                auto_generate_title = true,
                title_generation_opts = {
                    adapter = nil,
                    model = nil,
                    refresh_every_n_prompts = 0,
                    max_refreshes = 3,
                    format_title = function(original_title) return original_title end
                },
                summary = {
                    create_summary_keymap = "gcs",
                    browse_summaries_keymap = "gbs",
                    generation_opts = {
                        adapter = nil,
                        model = nil,
                        context_size = 90000,
                        include_references = true,
                        include_tool_outputs = true,
                        system_prompt = nil,
                        format_summary = nil,
                    },
                },
                memory = {
                    auto_create_memories_on_summary_generation = true,
                    vectorcode_exe = "vectorcode",
                    tool_opts = { default_num = 10 },
                    notify = true,
                    index_on_startup = false,
                },
            }
        }
    }
})

Filter chats using chat_filter
main
You can provide a chat_filter function in your configuration to narrow down which chats appear in the history browser. The function receives chat_data (of type ChatIndexData) and should return a boolean.
Examples:
- Filter by current working directory.
- Filter by recent chats (e.g., last 7 days).
```
-- Filter by current working directory
chat_filter = function(chat_data)
    return chat_data.cwd == vim.fn.getcwd()
end

-- Recent chats only (last 7 days)
chat_filter = function(chat_data)
    local seven_days_ago = os.time() - (7 * 24 * 60 * 60)
    return chat_data.updated_at >= seven_days_ago
end
```
https://github.com/ravitemer/codecompanion-history.nvim/blob/main/README.md
Use the CodeCompanion History API
main
The history extension exports functions via require("codecompanion").extensions.history. These allow programmatic management of chats and summaries.
Chat Management
- get_location(): Returns the storage path.
- save_chat(chat): Saves the provided chat (or the last chat if none provided).
- browse_chats(filter_fn): Opens the history browser with a custom filter.
- get_chats(filter_fn): Returns a table of all saved chat metadata.
- load_chat(save_id): Loads a specific chat.
- delete_chat(save_id): Deletes a chat.
- duplicate_chat(save_id, new_title): Duplicates a chat.
Summary Management
- generate_summary(chat): Generates a summary for the current chat.
- delete_summary(summary_id): Deletes a specific summary.
- get_summaries(): Returns a table of all summaries.
- load_summary(summary_id): Loads summary content.
```
local history = require("codecompanion").extensions.history

-- Browse chats with project filter
history.browse_chats(function(chat_data)
    return chat_data.project_root == utils.find_project_root()
end)

-- Get all saved chats metadata
local chats = history.get_chats()
local chat_data = history.load_chat("some_save_id")
history.delete_chat("some_save_id")

-- Duplicate a chat with custom title
local new_save_id = history.duplicate_chat("some_save_id", "My Custom Copy")

-- Summary operations
history.generate_summary() -- generates for current chat
local summaries = history.get_summaries()
local summary_content = history.load_summary("some_save_id")
```
https://github.com/ravitemer/codecompanion-history.nvim/blob/main/README.md

Reference ChatIndexData structure

main

When using chat_filter, browse_chats, or get_chats, the chat_data object follows this structure:

Field	Type	Description
`save_id`	string	Unique chat identifier (Unix timestamp)
`title`	string	Chat title
`cwd`	string	Working directory when saved
`project_root`	string	Detected project root
`adapter`	string	LLM adapter used
`model`	string	Model name
`updated_at`	number	Unix timestamp of last update
`message_count`	number	Number of messages in chat
`token_estimate`	number	Estimated token count

codecompanion-history.nvim

What's inside codecompanion-history.nvim

Commands

Chat Buffer Keymaps (Default)

History Browser Actions

Summary Browser Actions

Requirements

Key Configuration Areas

Chat Management

Summary System

Memory System (Requires VectorCode CLI)

Chat Management

Summary Management