feat: Add OpenAI agents config

sources/agents/agents.qrc

@@ -10,8 +10,12 @@
         <file>claude_quick_refactor_fast.toml</file>
         <file>openai_base_chat.toml</file>
         <file>openai_chat.toml</file>
+        <file>openai_chat_mini.toml</file>
         <file>openai_base_responses.toml</file>
         <file>openai_chat_responses.toml</file>
+        <file>openai_completion.toml</file>
+        <file>openai_compression.toml</file>
+        <file>openai_quick_refactor.toml</file>
         <file>google_base_chat.toml</file>
         <file>google_chat.toml</file>
         <file>mistral_base_chat.toml</file>

sources/agents/openai_chat.toml

@@ -2,9 +2,9 @@ schema_version = 1
 
 extends     = "OpenAI Base Chat"
 name        = "OpenAI Chat"
-description = "OpenAI GPT-4o — coding chat via Chat Completions."
+description = "OpenAI GPT-5.5 — flagship coding chat via Chat Completions at high reasoning effort. The default OpenAI choice. NOTE: GPT-5 models reject 'max_tokens' and 'temperature' on Chat Completions — use 'max_completion_tokens' + 'reasoning_effort'. Reasoning runs internally and is NOT surfaced as a thinking block on Chat Completions; use 'OpenAI Chat — Responses' for visible reasoning summaries."
 
-model           = "gpt-4o"
+model           = "gpt-5.5"
 enable_tools    = true
 enable_thinking = true
 tags            = ["chat", "openai", "cloud"]
@@ -12,5 +12,5 @@ tags            = ["chat", "openai", "cloud"]
 system_prompt = """{{ read_file(":/roles/qt-cpp-developer.md") }}"""
 
 [body]
-max_tokens  = 8192
+max_completion_tokens = 32000
-temperature = 0.7
+reasoning_effort      = "high"

sources/agents/openai_chat_mini.toml

@@ -0,0 +1,16 @@
+schema_version = 1
+
+extends     = "OpenAI Base Chat"
+name        = "OpenAI Chat — Mini"
+description = "OpenAI GPT-5.4 mini — fast, lower-cost coding chat via Chat Completions at medium reasoning effort. NOTE: GPT-5 models use 'max_completion_tokens' + 'reasoning_effort' and reject 'temperature' on Chat Completions."
+
+model           = "gpt-5.4-mini"
+enable_tools    = true
+enable_thinking = true
+tags            = ["chat", "openai", "cloud"]
+
+system_prompt = """{{ read_file(":/roles/qt-cpp-developer.md") }}"""
+
+[body]
+max_completion_tokens = 16000
+reasoning_effort      = "medium"

sources/agents/openai_chat_responses.toml

@@ -2,9 +2,9 @@ schema_version = 1
 
 extends     = "OpenAI Base Responses"
 name        = "OpenAI Chat — Responses"
-description = "OpenAI o4-mini — reasoning coding chat via the Responses API."
+description = "OpenAI GPT-5.1 — reasoning coding chat via the Responses API. Best for agentic coding; surfaces a reasoning summary (summary=auto) as a thinking block. Uses max_output_tokens + the reasoning object (reasoning models reject temperature)."
 
-model           = "o4-mini"
+model           = "gpt-5.1"
 enable_tools    = true
 enable_thinking = true
 tags            = ["chat", "openai", "responses", "cloud"]
@@ -12,5 +12,5 @@ tags            = ["chat", "openai", "responses", "cloud"]
 system_prompt = """{{ read_file(":/roles/qt-cpp-developer.md") }}"""
 
 [body]
-max_output_tokens = 25000
+max_output_tokens = 32000
-reasoning         = { effort = "medium", summary = "auto" }
+reasoning         = { effort = "high", summary = "auto" }

sources/agents/openai_completion.toml

@@ -0,0 +1,30 @@
+schema_version = 1
+
+extends     = "OpenAI Base Chat"
+name        = "OpenAI Completion"
+description = "OpenAI GPT-5.4 mini — code completion using the <code_context> chat format over Chat Completions. reasoning_effort=none for low latency. GPT-5 models reject 'temperature' AND 'stop' on Chat Completions, so completion cannot be pinned to t=0 and has no stop sequence — it relies on the prompt + max_completion_tokens to stay short."
+
+model = "gpt-5.4-mini"
+tags  = ["completion", "openai", "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]
+max_completion_tokens = 1024
+reasoning_effort      = "none"
+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>") }}
+  }
+]
+"""

sources/agents/openai_compression.toml

@@ -0,0 +1,15 @@
+schema_version = 1
+
+extends     = "OpenAI Base Chat"
+name        = "OpenAI Compression"
+description = "OpenAI GPT-5.4 mini — fast, low-cost conversation summarization for shorter chats. Carries the summary system prompt; no tools, reasoning_effort=none. GPT-5 models reject 'temperature' on Chat Completions."
+
+model        = "gpt-5.4-mini"
+enable_tools = false
+tags         = ["compression", "openai", "cloud"]
+
+system_prompt = """{{ read_file(":/tasks/chat-compressor.md") }}"""
+
+[body]
+max_completion_tokens = 16000
+reasoning_effort      = "none"

sources/agents/openai_quick_refactor.toml

@@ -0,0 +1,16 @@
+schema_version = 1
+
+extends     = "OpenAI Base Chat"
+name        = "OpenAI Quick Refactor"
+description = "OpenAI GPT-5.1 — agentic inline refactor with tools at high reasoning effort (gathers context before editing). Static output rules from :/tasks/quick-refactor.md; QuickRefactorHandler injects the live editor context (file, code, cursor/selection). GPT-5 models use 'max_completion_tokens' + 'reasoning_effort' and reject 'temperature'."
+
+model           = "gpt-5.1"
+enable_tools    = true
+enable_thinking = true
+tags            = ["refactor", "openai", "cloud"]
+
+system_prompt = """{{ read_file(":/tasks/quick-refactor.md") }}"""
+
+[body]
+max_completion_tokens = 16000
+reasoning_effort      = "high"