schema_version = 1

extends     = "Google Base Chat"
name        = "Google Completion"
description = "Google Gemini 3.1 Flash-Lite — code completion using the <code_context> chat format over generateContent. Thinking disabled (thinkingBudget=0) and temperature=0 for fast, deterministic insertions; stops at </code_context>."

model = "gemini-3.1-flash-lite"
tags  = ["completion", "gemini", "google", "cloud"]

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]
contents = """
[
  {
    "role": "user",
    "parts": [ { "text": {{ tojson("Here is the code context with insertion points:\\n<code_context>\\n" + ctx.prefix + "<cursor>" + ctx.suffix + "\\n</code_context>") }} } ]
  }
]
"""

[body.generationConfig]
maxOutputTokens = 1024
temperature     = 0
stopSequences   = ["</code_context>"]
thinkingConfig  = { thinkingBudget = 0 }