mirror of
https://github.com/Palm1r/QodeAssist.git
synced 2026-06-30 01:59:11 -04:00
32 lines
1.3 KiB
TOML
32 lines
1.3 KiB
TOML
schema_version = 1
|
|
|
|
extends = "Ollama Base Chat"
|
|
name = "Ollama Completion — Chat-style"
|
|
description = "Code completion via a chat request — the cursor sits in a <code_context> message on /api/chat, so the model just continues the code. Works with ANY normal chat/instruct model (qwen2.5-coder, qwen3, llama3, gemma, mistral, codellama-instruct, …), so use this when your model is NOT a dedicated FIM model. Tradeoff vs native 'Ollama Completion — FIM': a bit slower and may leak code fences or explanations, so slightly less precise."
|
|
|
|
model = "qwen2.5-coder:7b"
|
|
tags = ["completion", "ollama", "local", "fim", "8gb"]
|
|
|
|
system_prompt = """
|
|
{%- if language == "qml" %}{{ read_file(":/roles/code-completion-qml.md") }}
|
|
{%- else if language == "c-like" %}{{ read_file(":/roles/code-completion-c-like.md") }}
|
|
{%- else %}{{ read_file(":/roles/code-completion.md") }}
|
|
{%- endif %}
|
|
{{ read_file(":/tasks/code-completion.md") }}"""
|
|
|
|
[body]
|
|
messages = """
|
|
[
|
|
{% if existsIn(ctx, "system_prompt") %}
|
|
{ "role": "system", "content": {{ tojson(ctx.system_prompt) }} },
|
|
{% endif %}
|
|
{ "role": "user", "content": {{ tojson("Here is the code context with insertion points:\\n<code_context>\\n" + ctx.prefix + "<cursor>" + ctx.suffix + "\\n</code_context>") }} }
|
|
]
|
|
"""
|
|
|
|
[body.options]
|
|
num_predict = 256
|
|
temperature = 0.2
|
|
keep_alive = "5m"
|
|
stop = ["</code_context>"]
|