[Bug]: TypeError: Completions.create() got an unexpected keyword argument 'system' in chat_completions.py via _build_kwargs_from_profile

[rdxhemadri]

Bug Description

When running Hermes Agent with OpenAI-compatible third-party endpoints (e.g., OpenRouter and SiliconFlow Custom providers), the conversation loop intermittently crashes with a non-retryable client error. The runtime executes Completions.create() (the legacy text completions method or an improperly bound endpoint mapping) while explicitly forwarding a top-level system prompt keyword parameter. This causes a TypeError in the underlying OpenAI Python SDK wrapper. Because the error is caught as a non-retryable client validation exception, it propagates entirely through the fallback chain and triggers an unrecoverable infinite loop during agent state recovery.

Steps to Reproduce

1. Configure a custom OpenAI/OpenRouter provider profile (/opt/data/profiles/robin/profile.yaml).
2. Set up models using chat interfaces that utilize custom endpoint definitions (e.g., tencent/Hy3 via SiliconFlow or nvidia/nemotron-3-super-120b-a12b:free via OpenRouter).
3. Interact with the agent over an extended multi-turn session until the active context length escalates (e.g., ~31 messages, ~43,996 tokens).
4. Watch the agent crash on payload construction immediately during a fallback rotation or session continuation query.

Expected Behavior

The agent should construct a valid modern payload targetting the chat interface (ChatCompletions.create()) or map system variables safely inside standard messages arrays rather than appending system as a top-level kwarg payload parameter to the legacy Completions interface.

Actual Behavior

WARNING agent.conversation_loop: API call failed (attempt 1/3) error_type=TypeError thread=hermes-gateway_0:130297385428672 provider=custom base_url=https://siliconflow.com model=tencent/Hy3 summary=Completions.create() got an unexpected keyword argument 'system'

WARNING agent.conversation_loop: API call failed (attempt 1/3) error_type=TypeError thread=hermes-gateway_0:130297385428672 provider=openrouter base_url=https://openrouter.ai model=nvidia/nemotron-3-super-120b-a12b:free summary=Completions.create() got an unexpected keyword argument 'system'

ERROR agent.conversation_loop: Non-retryable client error: Completions.create() got an unexpected keyword argument 'system'

Affected Component

Agent Core (conversation loop, context compression, memory)

Messaging Platform (if gateway-related)

No response

Debug Report

"description": "Deep structured memory with algebraic reasoning. Use alongside the memory tool — memory for always-on context, fact_store for deep recall and compositional queries.\n\nACTIONS (simple → powerful):\n• add — Store a fact the user would expect you to remember.\n• search — Keyword lookup ('editor config', 'deploy process').\n• probe — Entity recall: ALL facts about a person/thing.\n• related — What connects to an entity? Structural adjacency.\n• reason — Compositional: facts connected to MULTIPLE entities simultaneously.\n• contradict — Memory hygiene: find facts making conflicting claims.\n• update/remove/list — CRUD operations.\n\nIMPORTANT: Before answering questions about the user, ALWAYS probe or reason first.",
            "parameters": {
              "type": "object",
              "properties": {
                "action": {
                  "type": "string",
                  "enum": [
                    "add",
                    "search",
                    "probe",
                    "related",
                    "reason",
                    "contradict",
                    "update",
                    "remove",
                    "list"
                  ]
                },
                "content": {
                  "type": "string",
                  "description": "Fact content (required for 'add')."
                },
                "query": {
                  "type": "string",
                  "description": "Search query (required for 'search')."
                },
                "entity": {
                  "type": "string",
                  "description": "Entity name for 'probe'/'related'."
                },
                "entities": {
                  "type": "array",
                  "items": {
                    "type": "string"
                  },
                  "description": "Entity names for 'reason'."
                },
                "fact_id": {
                  "type": "integer",
                  "description": "Fact ID for 'update'/'remove'."
                },
                "category": {
                  "type": "string",
                  "enum": [
                    "user_pref",
                    "project",
                    "tool",
                    "general"
                  ]
                },
                "tags": {
                  "type": "string",
                  "description": "Comma-separated tags."
                },
                "trust_delta": {
                  "type": "number",
                  "description": "Trust adjustment for 'update'."
                },
                "min_trust": {
                  "type": "number",
                  "description": "Minimum trust filter (default: 0.3)."
                },
                "limit": {
                  "type": "integer",
                  "description": "Max results (default: 10)."
                }
              },
              "required": [
                "action"
              ]
            }
          }
        },
        {
          "type": "function",
          "function": {
            "name": "fact_feedback",
            "description": "Rate a fact after using it. Mark 'helpful' if accurate, 'unhelpful' if outdated. This trains the memory — good facts rise, bad facts sink.",
            "parameters": {
              "type": "object",
              "properties": {
                "action": {
                  "type": "string",
                  "enum": [
                    "helpful",
                    "unhelpful"
                  ]
                },
                "fact_id": {
                  "type": "integer",
                  "description": "The fact ID to rate."
                }
              },
              "required": [
                "action",
                "fact_id"
              ]
            }
          }
        }
      ],
      "extra_body": {
        "session_id": "20260708_090850_04622636"
      },
      "system": "\n\n⚡ SKILL LOADING REMINDER: No domain-specific skills have been loaded recently (turn 17, freshness window: 15 turns). Before fixing, debugging, or modifying any system, you MUST load the relevant skill(s) using skill_view(). This is NON-NEGOTIABLE.\n\nDomain skill chains:\n[backtest] position-sizing → purged-cross-validation → overfitting-defense | [cron] cron-triage-protocol → kaggle-pipeline | [debug] systematic-debugging → skill-driven-debugging → pipeline-mastery | [feature] nse-market-microstructure → microstructure-features → fractional-differentiation | [hermes] hermes-agent | [kaggle] kaggle-env-compat → kaggle-api-behavior → pipeline-mastery → kaggle-pipeline"
    }
  },
  "error": {
    "type": "TypeError",
    "message": "Completions.create() got an unexpected keyword argument 'system'"
  }
}

Operating System

Linux 6.6.114.1-microsoft-standard-WSL2 x86_64

Python Version

No response

Hermes Version

No response

Additional Logs / Traceback (optional)

Root Cause Analysis (optional)

The error stems from agent/transports/chat_completions.py within _build_kwargs_from_profile():Around lines 570–600, dictionary iteration logic mapping user environment configurations assigns raw parameters dynamically: api_kwargs[k] = v.When it captures a root-level profile rule mapping (such as a system template variable or explicit model option named system), it appends it directly as an argument to the request block, exploding during unpacked instantiation: Completions.create(**api_kwargs)

Proposed Fix (optional)

No response

Are you willing to submit a PR for this?

• I'd like to fix this myself and submit a PR

[rdxhemadri]

do anyone try to open PR on this issue pls

[liuhao1024]

I investigated this issue and found that:

1. The error occurs when a parameter named is passed directly in (top-level kwargs to the OpenAI client).

2. In , the method collects from and unconditionally merges it into at line 540:

3. However, none of the existing provider profiles return a key in their . I checked all provider profiles (OpenRouter, Custom, DeepSeek, Qwen, Nous, etc.) — none of them emit as a top-level kwarg.

4. The only place a kwarg would come from is **user-configured **. In line 567-573 of :

5. The issue report suggests the problem is in , but that function doesn't directly handle profile rule mappings for . The actual root cause is likely in the config parsing layer that populates .

To proceed with a fix, I need to know:

• Is there a specific config file or code path that's adding to ?
• Can you share the relevant profile.yaml or config snippet that triggers this?

If the issue is indeed that is incorrectly being passed through , the fix would be to filter out known-invalid OpenAI API kwargs before they reach the client. However, I want to confirm the root cause first to avoid a blind fix.

[rdxhemadri]

config.yaml # this is the main config.yaml file

config.yaml # this is the profile config.yaml file

what else do you need

[rdxhemadri]

@teknium1 @liuhao1024 @LeonSGP43 pls help to fix it

[rdxhemadri]

this is debug report https://paste.rs/qPc1d