d4rkninja/tanpo-ops

Tanpo Ops

A compact operations specialist (~1.2B) for capacity planning, identity guardrails, OKR execution, vendor operations, handoffs, SOPs, operating cadence, and incident postmortems — built for local and inexpensive deployment.

Creator: d4rkninja Collection: Tanpo — Domain Specialists

Original upstream: LiquidAI/LFM2.5-1.2B-Instruct (~1.17B parameters, 32,768-token context, designed for edge/on-device deployment).

Fine-tuning: Unsloth-compatible loading of that checkpoint via hub id `unsloth/LFM2.5-1.2B-Instruct` (LoRA / PEFT).

This repository hosts the merged Transformers weights (LoRA merged into the base).

Overview

Tanpo is a family of compact domain-specialized business models for local / edge / inexpensive deployment. Different specialists cover different workflows. One compact architecture (LiquidAI/LFM2.5-1.2B-Instruct) → multiple focused specialists → each ships Full/Merged | LoRA | GGUF. Tanpo Ops is one specialist in that family (not a frontier or general-purpose model).

Related artifacts:

• —LoRA adapter: d4rkninja/tanpo-ops-LoRA
• —GGUF quants: d4rkninja/tanpo-ops-GGUF — prefer Q4_K_M when available

Best For

• —Capacity-planning drafts, resourcing assumptions, and operational tradeoff analysis
• —OKR decomposition, execution plans, status reviews, and operating cadence
• —Vendor-operations checklists, handoff design, and SOP drafting
• —Incident-postmortem structure, follow-up actions, and identity-aware operational workflows

Not Designed For

• —Fully automated staffing, access, vendor, incident-severity, or other consequential operational decisions
• —Fabricating operational facts, impersonation, bypassing identity controls, or unauthorized access
• —Guarantees about capacity, uptime, compliance, incident outcomes, or business performance
• —General coding or non-operations chat

Why a Specialist Model?

Operations work rewards repeatable structure (inputs, assumptions, owner, timing, dependencies, risks, decision, and next step). Specializing a small model for those workflows enables private, low-cost inference without a large general model.

Evaluation

Internal automated domain evaluation (DarkLab harness). Treat as directional, not an industry benchmark.

Largest gains were on capacity_planning (+5.5 percentage points), identity (+17.1 percentage points), okr_execution (+11.1 percentage points), and vendor_ops (+5.6 percentage points).

Trailed on handoffs (−7.4 percentage points) and on sops, operating_cadence, and incident_postmortem (~−3.7 percentage points each). Overall still ahead of base; these categories are disclosed.

Artifacts: `evaluation/` — `COMPARE_OPS.md`.

Methodology: DarkLab automated domain evaluation. Same prompts and generation config for base vs fine-tune. Not an industry benchmark.

Limitations of this eval: Automated rubrics can reward structure over real-world quality; sample size is small; results may not transfer outside the task distribution.

Example Prompts

1. 1.User: Build a capacity plan from these workload assumptions, separating facts, assumptions, constraints, risks, and decisions needed.
2. 2.User: Turn this quarterly objective into measurable OKRs with owners, dependencies, milestones, and a weekly operating cadence.
3. 3.User: Draft an incident postmortem outline from these notes; do not invent impact, root cause, or remediation facts.

Inference

from transformers import AutoModelForCausalLM, AutoTokenizer

repo = "d4rkninja/tanpo-ops"
tokenizer = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
    repo,
    torch_dtype="auto",
    device_map="auto"
)

messages = [
    {"role": "system", "content": "You are Tanpo Ops, a practical operations assistant."},
    {"role": "user", "content": "Turn this quarterly objective into measurable OKRs with owners, dependencies, milestones, and a weekly operating cadence."},
]
inputs = tokenizer.apply_chat_template(
    messages, add_generation_prompt=True, return_tensors="pt"
).to(model.device)
out = model.generate(inputs, max_new_tokens=512)
print(tokenizer.decode(out[0][inputs.shape[-1]:], skip_special_tokens=True))

Training

Verified from published adapter configs / training artifacts (no unverified hyperparams):

Merged via PEFT merge_and_unload into full weights in this repo.

Dataset

• —d4rkninja/tanpo-ops-sft — format-fixed chat SFT examples (documented on the dataset card).

Limitations

• —Specialized: quality drops outside the operations and business-workflow distribution.
• —~1.2B scale: limited world knowledge and long-horizon reasoning vs larger models.
• —Capacity, staffing, vendor, identity, OKR, and incident recommendations can be wrong or incomplete; verify against source systems, policies, and operational context.
• —Eval gains are rubric-based and directional only; handoffs, sops, operating_cadence, and incident_postmortem still trail the base.

Responsible Use

Not a substitute for operational judgment, source-system data, access-control policy, security review, incident command, or legal/compliance advice. Humans must review capacity plans, staffing recommendations, access-sensitive workflows, vendor decisions, OKRs, SOPs, and incident records before action. Do not use for deception, harassment, unauthorized access, or discriminatory treatment.

License

license: other / license_name: lfm-1.0

Tanpo merged and GGUF weights are derivatives of LiquidAI/LFM2.5-1.2B-Instruct under the LFM Open License v1.0 (including the commercial Threshold of approximately $10M annual revenue). See the base model card and its LICENSE file. Credit: LiquidAI. Do not treat this stack as Apache-2.0.

Tanpo Family

Tanpo is a family of compact domain-specialized models for focused business workflows.

This specialist is available as:

• —Merged: `d4rkninja/tanpo-ops`
• —LoRA: `d4rkninja/tanpo-ops-LoRA`
• —GGUF: `d4rkninja/tanpo-ops-GGUF`

Browse all Tanpo specialists: Tanpo — Domain Specialists