Ollama streaming response fail with "attempt to call method 'match' (a nil value)" error

[ericyeon34]

Bug Description

When using gp.nvim with Ollama as a provider, streaming responses fail with a Lua error at dispatcher.lua:328. The plugin is unable to parse Ollama's Server-Sent Events (SSE) response format.

Error Message

Error executing callback:
...nvim-data/lazy/gp.nvim/lua/gp/dispatcher.lua:328: attempt to call method 'match' (a nil value)
stack traceback:
        ...nvim-data/lazy/gp.nvim/lua/gp/dispatcher.lua:328: in function 'process_lines'
        ...nvim-data/lazy/gp.nvim/lua/gp/dispatcher.lua:362: in function 'out_reader'
        ...nvim-data/lazy/gp.nvim/lua/gp/tasker.lua:166: in function <...nvim-data/lazy/gp.nvim/lua/gp/tasker.lua:158>

Additionally:

Gp.nvim: ollama response is empty:
'data: {"id":"chatcmpl-223","object":"chat.completion.chunk","created":1770957764,"model":"qwen2.5-coder:14b",...}'

Configuration

{
    "robitx/gp.nvim",
    config = function()
        require("gp").setup({
            providers = {
                openai = {
                    disable = true,
                },
                ollama = {
                    endpoint = "http://localhost:11434/v1/chat/completions",
                    secret = "ollama",
                },
            },
            agents = {
                {
                    name = "Clanker",
                    provider = "ollama",
                    chat = true,
                    command = true,
                    model = { model = "qwen2.5-coder:14b", temperature = 0.2, top_p = 1 },
                    system_prompt = "...",
                },
            },
            default_chat_agent = "Clanker",
            default_command_agent = "Clanker",
        })
    end,
}

Root Cause

Ollama's OpenAI-compatible API returns streaming responses in SSE format with a data: prefix:

data: {"id":"chatcmpl-223","object":"chat.completion.chunk",...}

At line 328 in dispatcher.lua, the code attempts to call line:match() without checking if line is a string or handling the data: prefix. This causes the error when processing Ollama's response format.

Current Code (line 328)

if qt.provider == "ollama" then
    if line:match('"message":') and line:match('"content":') then
        local success, decoded = pcall(vim.json.decode, line)
        if success and decoded.message and decoded.message.content then
            content = decoded.message.content
        end
    end
end

Proposed Fix

if qt.provider == "ollama" then
    -- Strip "data: " prefix that Ollama adds to SSE streams
    if type(line) == "string" then
        line = line:gsub("^data: ", "")
    end
    
    if type(line) == "string" and line:match('"message":') and line:match('"content":') then
        local success, decoded = pcall(vim.json.decode, line)
        if success and decoded.message and decoded.message.content then
            content = decoded.message.content
        end
    end
end

Environment

• Neovim version: v0.11.6
• gp.nvim version: latest (main branch)
• Ollama version: 0.16.0
• OS: Windows 11

Steps to Reproduce

1. Install and configure Ollama with any model (e.g., qwen2.5-coder:14b)
2. Configure gp.nvim with Ollama provider as shown above
3. Attempt to use :GpChatNew and send a message
4. Observe the error

Expected Behavior

Ollama streaming responses should be parsed correctly and displayed in the chat buffer.

Actual Behavior

Error is thrown and response shows as "empty" despite Ollama successfully returning data.

Workaround

Manually edit dispatcher.lua line 328 with the proposed fix above.