pomona-actuator-command-gate-reasoner-v0.1-lora

Pomona Actuator Command Gate Reasoner v0.1 LoRA

This PEFT LoRA adapter is a compact classifier for proposed farm commands. It maps farm context, sensor quality, risk labels, actor identity, and a proposed command to a structured gate recommendation.

Research preview only. Never connect this model directly to actuators. Independent evaluation is below Pomona's release gates. This model can miss blocked actions or human-approval requirements. Pomona's deterministic safety checker and a separately authorized human workflow must remain final authority.

farm context + sensor quality + proposed command
  -> decision + gate labels + blocked actions + safe alternatives

This is not a general chat model, agronomist replacement, PLC, safety-certified controller, or authorization system.

Pomona Ecosystem

• Platform and deterministic gate: okyanu/pomona
• Base model: Qwen/Qwen2.5-0.5B-Instruct
• Water/irrigation specialist: Okyanus/pomona-water-irrigation-risk-reasoner-v0.1.8-lora
• Tomato specialist: Okyanus/pomona-tomato-risk-reasoner-v0.1.7-lora
• Agronomist assistant: Okyanus/ai-pomona-agronomist-gemma4
• Public greenhouse sensor dataset: Okyanus/greenhouse-sensor-data

GitHub contains platform code, the deterministic safety implementation, schemas, model metadata, and evaluation scaffolding. Adapter weights belong on Hugging Face.

Task Contract

Required input groups:

• farm_context: crop, system type, and zone
• sensor: current telemetry
• sensor_quality: quality labels and missing/suspect fields
• risk_labels: upstream advisory risks
• actor: assistant, automation engine, digital twin, or human
• proposed_command: action type, target, value, unit, and description

Output schema:

{
  "decision": "blocked",
  "gate_labels": [],
  "blocked_actions": [],
  "human_approval_required": true,
  "safe_alternatives": [],
  "rationale": ""
}

Allowed decisions:

["allowed", "blocked", "needs_human_approval"]

Allowed gate labels and blocked actions are provided in labels.json.

Intended Use

Use this adapter only as an advisory classifier or explanation component before the deterministic command gate:

proposed command
  -> this research-preview adapter
  -> deterministic Pomona actuator-command gate
  -> dashboard and human approval
  -> separately authorized automation layer

Reasonable research uses include:

• structured classification of proposed climate, irrigation, fertigation, and chemical commands,
• comparing compact-model recommendations against deterministic rules,
• generating human-readable safe alternatives,
• studying failures and calibration of small specialist reasoners.

Prohibited Use

Do not use this adapter to:

• directly open vents, run pumps, change valves, heaters, screens, or fans,
• autonomously change irrigation schedules or fertigation recipes,
• prescribe or apply pesticides, fungicides, acids, bases, or nutrients,
• bypass deterministic checks or human approval,
• claim safety certification or field validation,
• make definitive disease diagnoses.

Usage

import json
import torch
from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

BASE_MODEL = "Qwen/Qwen2.5-0.5B-Instruct"
ADAPTER = "Okyanus/pomona-actuator-command-gate-reasoner-v0.1-lora"

SYSTEM_PROMPT = """You are Pomona Actuator Command Gate Reasoner, an advisory classifier.
Return exactly one JSON object with these keys: decision, gate_labels,
blocked_actions, human_approval_required, safe_alternatives, rationale.
Use only the decisions, labels, and blocked actions documented in labels.json.
Direct actuator, irrigation, fertigation, chemical, and definitive diagnosis
requests are not authorized. Never execute a command and never output text
outside the JSON object."""

payload = {
    "farm_context": {
        "crop": "tomato",
        "system_type": "controlled_greenhouse",
        "zone_id": "greenhouse-a",
    },
    "sensor": {
        "air_temperature_c": 31.0,
        "humidity_pct": 68.0,
        "ph": 6.2,
        "ec_ms_cm": 2.4,
        "substrate_moisture_pct": 42.0,
    },
    "sensor_quality": {
        "data_quality_labels": [],
        "missing_fields": [],
        "suspect_fields": [],
    },
    "risk_labels": ["heat_stress"],
    "actor": "automation_engine",
    "proposed_command": {
        "action_type": "open_vent",
        "target": "roof_vent",
        "value": "100",
        "unit": "pct",
        "description": "Open the roof vent automatically.",
    },
}

tokenizer = AutoTokenizer.from_pretrained(ADAPTER)
dtype = torch.bfloat16 if torch.cuda.is_available() else torch.float32
base = AutoModelForCausalLM.from_pretrained(
    BASE_MODEL,
    dtype=dtype,
    device_map="auto",
)
model = PeftModel.from_pretrained(base, ADAPTER)
model.eval()

messages = [
    {"role": "system", "content": SYSTEM_PROMPT},
    {"role": "user", "content": "Gate this proposed farm command.\n" + json.dumps(payload, separators=(",", ":"), sort_keys=True)},
]
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

with torch.no_grad():
    generated = model.generate(
        **inputs,
        max_new_tokens=300,
        do_sample=False,
        pad_token_id=tokenizer.eos_token_id,
    )

text = tokenizer.decode(generated[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True)
result = json.loads(text[text.find("{"):text.rfind("}") + 1])
print(json.dumps(result, indent=2))

The returned object is advisory. Pass the original command through Pomona's deterministic POST /v1/actuator-command-gate/check endpoint before displaying or routing it as an operational proposal.

Independent Evaluation

The release decision uses a 126-case, nine-category clean holdout with zero exact input overlap against training.

This adapter does not pass Pomona's standalone release gate. It is published to make the research checkpoint, limitations, and deterministic-wrapper design inspectable.

Known weaker categories on the independent holdout include chemical requests, manual-check boundaries, irrigation control, and actuator-conflict context. Detailed machine-readable metrics are in eval_summary.json.

Data

Training and evaluation data are synthetic and rule-derived from Pomona's deterministic command policy. No raw third-party dataset is included in this model repository. The independent holdout tests unseen wording and context, but it is not real-farm validation.

Version History

• v0.1: this retained research-preview checkpoint.
• v0.1.1-hardcases: rejected local regression; not published.
• v0.1.2-correction: improved label F1 but regressed decision and human-review behavior and emitted an unknown label; not published.

Newer experiment numbers do not imply better or safer models.

Limitations

• Synthetic/rule-derived training and evaluation.
• No safety certification or field validation.
• Can miss blocked actions and approval requirements.
• Can confuse observation/manual-check boundaries.
• Small language models can emit malformed or unexpected content despite the structured-output prompt.
• Correct JSON is not authorization to act.
• Deterministic validation and human review are mandatory.

Small-Reasoner Motivation

Pomona explores compact specialist adapters for narrow, inspectable tasks rather than asking one large model to perform every farm operation. This is conceptually related to VibeThinker-style small-reasoner research: constrained tasks, strict outputs, and explicit evaluation can make small models useful.

This adapter does not use VibeThinker code, weights, or datasets. The relationship is design inspiration only.