d4rkninja/tanpo-product

Tanpo Product

A compact product-strategy specialist (~1.2B) for roadmap tradeoffs, prioritization, metrics, and CEO-level product thinking — 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. Tanpo Product is one specialist in that family (not a frontier or general-purpose model).

Related artifacts:

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

Best For

• —Kill / keep / invest and prioritization write-ups
• —Roadmap sequencing and scope cuts
• —North-star metrics, OKRs, and operating cadence
• —Competitive positioning, pricing/monetization frames, and discovery synthesis

Not Designed For

• —Substituting primary customer research or user interviews
• —Authoritative market sizing or financial forecasts
• —Legal, security, or compliance decisions
• —General coding or non-product knowledge work

Why a Specialist Model?

Product decisions need structured tradeoffs more than generic chat. Fine-tuning a small model on product-workflow formats makes edge deployment practical without a large general model.

Evaluation

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

Artifacts: `evaluation/` — `COMPARE_BASE.md`, `product_tasks.jsonl`, `score_rubric.md`, `evaluation/README.md`.

Methodology: DarkLab automated domain evaluation on the product training/eval format revision fine-tune (~20 held-out tasks). 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: We have three bets for next quarter: (A) usage-based billing, (B) Salesforce sync, (C) mobile offline mode. Only engineering capacity for one. Recommend with kill criteria.
2. 2.User: Draft a one-page PRD outline for an in-app 'saved views' feature for a B2B analytics product. Include problem, users, success metric, and non-goals.
3. 3.User: Propose a north-star metric and 3 supporting input metrics for a vertical SaaS CRM for dental clinics.

Inference

from transformers import AutoModelForCausalLM, AutoTokenizer

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

messages = [
    {"role": "system", "content": 'You are Tanpo Product, a practical product-strategy assistant.'},
    {"role": "user", "content": 'We have three bets for next quarter: (A) usage-based billing, (B) Salesforce sync, (C) mobile offline mode. Only engineering capacity for one. Recommend with kill criteria.'},
]
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-product-sft — ~10k format-fixed chat SFT examples (documented on the dataset card).

Limitations

• —Specialized: quality drops outside the product workflow distribution.
• —~1.2B scale: limited world knowledge and long-horizon reasoning vs larger models.
• —Can sound decisive when evidence is thin — require human judgment for consequential product bets.
• —Eval gains are rubric-based and directional only.

Responsible Use

Not a substitute for customer research, board fiduciary judgment, or professional analysis. Treat outputs as drafts for human product leaders.

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-product`
• —LoRA: `d4rkninja/tanpo-product-LoRA`
• —GGUF: `d4rkninja/tanpo-product-GGUF`

Browse all Tanpo specialists: Tanpo — Domain Specialists