fix: preserve GPT frontier model roles

README.md

@@ -116,10 +116,11 @@ The checked-in default CLI and web profiles use these core components:
 - [Tavily Search API](https://tavily.com/) for web search
 - Serper, SerpAPI, or SearchAPI for Google Scholar paper search
 
-The shipped frontier profile, `configs/config_frontier_models.yml`, uses GPT-5.6 Sol for deep-research orchestration,
-planning, and writing, and GPT-5.6 Luna for source routing and research. Treat any bring-your-own model or modified
-profile as an experimental customization until the complete workflow is evaluated with that exact model, prompt,
-hyperparameter, tool-calling, and structured-output configuration. Refer to
+The shipped frontier profile, `configs/config_frontier_models.yml`, uses GPT-5.6 Luna for intent classification,
+shallow research, source routing, and deep-research execution, with GPT-5.6 Sol for clarification, orchestration,
+planning, and writing. Treat any bring-your-own model or modified profile as an experimental customization until the
+complete workflow is evaluated with that exact model, prompt, hyperparameter, tool-calling, and structured-output
+configuration. Refer to
 [Configuration Files](#configuration-files); there is no single all-features profile.
 
 ## Target Audience
@@ -294,7 +295,7 @@ The `configs/` directory holds YAML workflow configs that define agents, tools,
 | `config_web_frag.yml` | Nemotron 3.5 Lightning; Nemotron 3 Ultra | Web/API and Helm base with Foundational RAG plus Tavily. Requires separately deployed RAG query and ingestion services. |
 | `config_web_opensearch.yml` | Nemotron 3.5 Lightning; Nemotron 3 Ultra; Nemotron 3 Embed | Web/API with built-in OpenSearch knowledge retrieval plus Tavily; supports self-hosted, `es`, and `aoss` authentication modes. |
 | `config_web_azure_ai_search.yml` | Nemotron 3.5 Lightning; Nemotron 3 Ultra; Nemotron 3 Embed | Web/API with Azure AI Search knowledge retrieval plus Tavily; supports API-key and Azure identity authentication. |
-| `config_frontier_models.yml` | GPT Sol/Luna; Nemotron 3.5 Lightning; Gemma 4 summary | LlamaIndex profile using GPT Sol/Luna for deep research and Nemotron 3.5 Lightning for intent/shallow. Requires `OPENAI_API_KEY` and `NVIDIA_API_KEY`. |
+| `config_frontier_models.yml` | GPT Sol/Luna; Gemma 4 summary | LlamaIndex frontier profile using GPT Luna for intent, shallow research, source routing, and research, with GPT Sol for clarification, orchestration, planning, and writing. Requires `OPENAI_API_KEY` and `NVIDIA_API_KEY`. |
 | `config_web_default_guardrails.yml` | Nemotron 3.5 Lightning; Nemotron 3 Ultra; Gemma 4 summary | LlamaIndex profile with workflow Guardrails explicitly attached, shallow-agent Guardrails dynamically attached through `workflow_functions`, and async deep-agent Guardrails applied by the AI-Q runner from the same target configuration. |
 | `config_web_frag_mcp_auth.yml` | Nemotron 3.5 Lightning; Nemotron 3 Ultra | Foundational RAG plus an opt-in protected per-user OAuth MCP source example. Requires a real MCP endpoint and shared token store. |
 | `config_domain_routing_and_skills.yml` | Nemotron 3 Ultra; Gemma 4 summary | Direct deep-research profile with domain routing, DuckDuckGo news, Polymarket, enabled Serper paper search, LlamaIndex, built-in skills, and a fresh per-job Modal sandbox. |

configs/config_frontier_models.yml

@@ -51,31 +51,21 @@ general:
         - "*"
 
 llms:
-  nemotron_lightning_intent_llm:
-    _type: nim
-    model_name: nvidia/nemotron-3.5-lightning-30b-a3b
-    base_url: "https://integrate.api.nvidia.com/v1"
-    api_key: ${NVIDIA_API_KEY}
-    temperature: 0.1
-    top_p: 0.9
+  gpt_luna_intent_llm:
+    _type: openai
+    model_name: "gpt-5.6-luna"
+    api_key: ${OPENAI_API_KEY}
     max_tokens: 1024
-    num_retries: 5
+    num_retries: 2
     parallel_tool_calls: false
-    chat_template_kwargs:
-      enable_thinking: false
 
-  nemotron_lightning_agent_llm:
-    _type: nim
-    model_name: nvidia/nemotron-3.5-lightning-30b-a3b
-    base_url: "https://integrate.api.nvidia.com/v1"
-    api_key: ${NVIDIA_API_KEY}
-    temperature: 0.2
-    top_p: 0.7
+  gpt_luna_shallow_llm:
+    _type: openai
+    model_name: "gpt-5.6-luna"
+    api_key: ${OPENAI_API_KEY}
     max_tokens: 8192
-    num_retries: 5
+    num_retries: 2
     parallel_tool_calls: false
-    chat_template_kwargs:
-      enable_thinking: true
 
   gpt_sol_agent_llm:
     _type: openai
@@ -141,7 +131,7 @@ functions:
 
   intent_classifier:
     _type: intent_classifier
-    llm: nemotron_lightning_intent_llm
+    llm: gpt_luna_intent_llm
     verbose: true
     tools:
       - web_search_tool
@@ -160,7 +150,7 @@ functions:
 
   shallow_research_agent:
     _type: shallow_research_agent
-    llm: nemotron_lightning_agent_llm
+    llm: gpt_luna_shallow_llm
     verbose: true
     tools:
       - web_search_tool

docs/notebooks/0_Getting_Started_with_AIQ.ipynb

@@ -418,7 +418,7 @@
     "**Main agents:** intent classifier, shallow research, deep research. \n",
     "**Tools:** web search tool for shallow researcher and advanced web search for deep research.\n",
     "\n",
-    ">**Optional — Frontier model config:** For improved deep research report quality, you can use `configs/config_frontier_models.yml`, which assigns `gpt-5.6-sol` to deep-research orchestration, planning, and writing, and `gpt-5.6-luna` to source routing and research. When using Docker Compose or the web UI, set `BACKEND_CONFIG` to `/app/configs/config_frontier_models.yml`. Remember to set the required API key in your `.env` file.\n"
+    ">**Optional — Frontier model config:** For improved research quality, you can use `configs/config_frontier_models.yml`, which assigns `gpt-5.6-luna` to intent classification, shallow research, source routing, and deep-research execution, and `gpt-5.6-sol` to clarification, orchestration, planning, and writing. When using Docker Compose or the web UI, set `BACKEND_CONFIG` to `/app/configs/config_frontier_models.yml`. Remember to set the required API keys in your `.env` file.\n"
    ]
   },
   {

docs/source/customization/configuration-reference.md

@@ -710,7 +710,7 @@ only the additional sections you need.
 | `configs/config_web_azure_ai_search.yml` | Web API | Azure AI Search knowledge retrieval and web search |
 | `configs/config_web_frag.yml` | Web API / Helm base | Foundational RAG plus Tavily. Requires separately deployed RAG query and ingestion services. Paper search is commented out. |
 | `configs/config_web_opensearch.yml` | Web API | Built-in OpenSearch knowledge backend plus Tavily. Supports unauthenticated or basic self-hosted OpenSearch and SigV4 (`es` or `aoss`); infrastructure and credentials are deployment opt-ins. |
-| `configs/config_frontier_models.yml` | Web API | Shipped LlamaIndex frontier profile: Nemotron 3.5 Lightning for intent/shallow, GPT-5.6 Sol for clarification/orchestration/planning/writing, GPT-5.6 Luna for source routing/research, and Gemma 4 for summaries. Requires `NVIDIA_API_KEY`, `OPENAI_API_KEY`, and `TAVILY_API_KEY` for the enabled Tavily tools; the commented paper-search opt-in requires `SERPER_API_KEY` when enabled. Validate the complete workflow against the configured provider endpoints before deployment. |
+| `configs/config_frontier_models.yml` | Web API | Shipped LlamaIndex frontier profile: GPT-5.6 Luna for intent/shallow/source routing/research, GPT-5.6 Sol for clarification/orchestration/planning/writing, and Gemma 4 for summaries. Requires `NVIDIA_API_KEY`, `OPENAI_API_KEY`, and `TAVILY_API_KEY` for the enabled Tavily tools; the commented paper-search opt-in requires `SERPER_API_KEY` when enabled. Validate the complete workflow against the configured provider endpoints before deployment. |
 | `configs/config_web_default_guardrails.yml` | Web API | LlamaIndex with workflow Guardrails attached explicitly, shallow-agent Guardrails dynamically attached through `workflow_functions`, and async deep-agent Guardrails applied by the AI-Q runner from the same target configuration. |
 | `configs/config_web_frag_mcp_auth.yml` | Web API | Foundational RAG plus a protected per-user OAuth MCP source example. Requires a real protected MCP endpoint and shared token-store configuration; it is not a zero-config default. |
 | `configs/config_domain_routing_and_skills.yml` | Direct deep-research workflow | Automatic domain routing, Tavily, DuckDuckGo news, Polymarket, LlamaIndex, enabled Serper paper search, built-in skills, and a Modal sandbox. Requires the corresponding service credentials and Modal setup. |

docs/source/customization/swapping-models.md

@@ -13,7 +13,7 @@ AI-Q 2.2 ships these exact model assignments and parameters:
 | Configuration | Intent and shallow roles | Clarification and deep-research roles | Optional summary role |
 | --- | --- | --- | --- |
 | `configs/config_cli_default.yml`, `configs/config_web_default_llamaindex.yml` | `nvidia/nemotron-3.5-lightning-30b-a3b` | `nvidia/nemotron-3-ultra-550b-a55b` for clarification, orchestration, source routing, research, planning, and writing | `google/gemma-4-31b-it` in the web profile |
-| `configs/config_frontier_models.yml` | `nvidia/nemotron-3.5-lightning-30b-a3b` | `gpt-5.6-sol` for clarification, orchestration, planning, and writing; `gpt-5.6-luna` for source routing and research | `google/gemma-4-31b-it` |
+| `configs/config_frontier_models.yml` | `gpt-5.6-luna` | `gpt-5.6-sol` for clarification, orchestration, planning, and writing; `gpt-5.6-luna` for source routing and research | `google/gemma-4-31b-it` |
 
 The checked-in files define the documented compatibility boundary; they are not a substitute for an end-to-end run
 against your provider endpoints and credentials. Changing a model, endpoint, role assignment, prompt, or inference

docs/source/examples/hybrid-frontier-model.md

@@ -5,14 +5,14 @@ SPDX-License-Identifier: Apache-2.0
 
 # Hybrid Frontier Model
 
-The checked-in `configs/config_frontier_models.yml` profile combines NVIDIA NIM models for intent classification and
-shallow research with GPT Sol/Luna for clarification and deep research.
+The checked-in `configs/config_frontier_models.yml` profile uses GPT Luna for bounded classification and research
+roles, GPT Sol for coordination and writing, and NVIDIA NIM for document summaries.
 
 ## Shipped Configuration
 
 | Role | Model |
 | --- | --- |
-| Intent classification and shallow research | `nvidia/nemotron-3.5-lightning-30b-a3b` |
+| Intent classification and shallow research | `gpt-5.6-luna` with role-specific token budgets |
 | Clarification, orchestration, and planning | `gpt-5.6-sol` |
 | Source routing and research | `gpt-5.6-luna` |
 | Report writing | `gpt-5.6-sol` with the writer token budget from the checked-in config |
@@ -31,7 +31,7 @@ as experimental until the complete workflow passes evaluation.
 
 ## Prerequisites
 
-- `NVIDIA_API_KEY` for NVIDIA NIM models
+- `NVIDIA_API_KEY` for the Gemma document-summary model
 - `OPENAI_API_KEY` for the GPT Sol/Luna roles
 - `TAVILY_API_KEY` for the default web-search tools
 

docs/source/examples/index.md

@@ -14,5 +14,5 @@ Complete, annotated configuration examples for common use cases.
 | [Full Pipeline -- Foundational RAG](./full-pipeline-web.md) | Complete production setup with hosted RAG | `config_web_frag.yml` |
 | [Azure AI Search Knowledge Layer](./azure-ai-search.md) | Managed hybrid document retrieval | `config_web_azure_ai_search.yml` |
 | [CLI with Local NIMs](./cli-with-local-nims.md) | Interactive CLI mode with self-hosted NIM models | `config_cli_default.yml` |
-| [Hybrid Frontier Model](./hybrid-frontier-model.md) | NIM for intent/shallow + shipped GPT Sol/Luna role assignment | `config_frontier_models.yml` |
+| [Hybrid Frontier Model](./hybrid-frontier-model.md) | GPT Luna for intent/research + GPT Sol for coordination/writing | `config_frontier_models.yml` |
 | [Deep Research Skills and Sandbox](./skills-sandbox/index.md) | DeepAgents skills with provider-backed sandbox execution for quantitative research workflows | `config_domain_routing_and_skills.yml` |

tests/aiq_agent/agents/deep_researcher/test_deepagents_runtime.py

@@ -50,6 +50,18 @@ def test_frontier_profile_uses_validated_gpt_role_split() -> None:
     config = yaml.safe_load(Path("configs/config_frontier_models.yml").read_text(encoding="utf-8"))
     llms = config["llms"]
 
+    assert llms["gpt_luna_intent_llm"] == {
+        "_type": "openai",
+        "model_name": "gpt-5.6-luna",
+        "api_key": "${OPENAI_API_KEY}",
+        "max_tokens": 1024,
+        "num_retries": 2,
+        "parallel_tool_calls": False,
+    }
+    assert llms["gpt_luna_shallow_llm"] == {
+        **llms["gpt_luna_intent_llm"],
+        "max_tokens": 8192,
+    }
     assert llms["gpt_sol_agent_llm"] == {
         "_type": "openai",
         "model_name": "gpt-5.6-sol",
@@ -66,6 +78,8 @@ def test_frontier_profile_uses_validated_gpt_role_split() -> None:
         **llms["gpt_sol_agent_llm"],
         "model_name": "gpt-5.6-luna",
     }
+    assert config["functions"]["intent_classifier"]["llm"] == "gpt_luna_intent_llm"
+    assert config["functions"]["shallow_research_agent"]["llm"] == "gpt_luna_shallow_llm"
     deep_research = config["functions"]["deep_research_agent"]
     assert {
         role: deep_research[f"{role}_llm"]

tests/aiq_agent/test_default_model_profiles.py

@@ -83,13 +83,16 @@ def _registered_source_tools(config: dict) -> set[str]:
 def test_default_profiles_use_role_appropriate_models(config_path: Path):
     config = _load_config(config_path)
     functions = config.get("functions", {})
+    is_frontier_profile = config_path.name == "config_frontier_models.yml"
 
     for function in functions.values():
         if not isinstance(function, dict):
             continue
 
         function_type = function.get("_type")
         if function_type == "intent_classifier":
+            if is_frontier_profile:
+                continue
             alias = function["llm"]
             assert alias == "nemotron_lightning_intent_llm"
             assert _model_for_alias(config, alias) == LIGHTNING_MODEL
@@ -101,6 +104,8 @@ def test_default_profiles_use_role_appropriate_models(config_path: Path):
             assert not config["llms"][alias]["parallel_tool_calls"]
             assert not _thinking_enabled(config, alias)
         elif function_type == "shallow_research_agent":
+            if is_frontier_profile:
+                continue
             alias = function["llm"]
             assert alias == "nemotron_lightning_agent_llm"
             assert _model_for_alias(config, alias) == LIGHTNING_MODEL
@@ -111,9 +116,9 @@ def test_default_profiles_use_role_appropriate_models(config_path: Path):
             assert config["llms"][alias]["max_tokens"] == 8192
             assert not config["llms"][alias]["parallel_tool_calls"]
             assert _thinking_enabled(config, alias)
-        elif config_path.name != "config_frontier_models.yml" and function_type == "clarifier_agent":
+        elif not is_frontier_profile and function_type == "clarifier_agent":
             assert _model_for_alias(config, function["llm"]) == ULTRA_MODEL
-        elif config_path.name != "config_frontier_models.yml" and function_type == "deep_research_agent":
+        elif not is_frontier_profile and function_type == "deep_research_agent":
             assert function["writer_llm"] == "nemotron_ultra_writer_llm"
             for role in (
                 "orchestrator_llm",