@pikaa-ai/pikaa 0.3.22 → 0.3.24

package/skills/transformers/SKILL.md

@@ -1,195 +0,0 @@
----
-name: transformers
-description: Hugging Face Transformers for loading Hub models, running pipeline inference, text generation, and Trainer fine-tuning on NLP, vision, audio, and multimodal tasks. Use when working with AutoModel, pipelines, tokenizers, or TrainingArguments—not for general ML outside the Transformers library.
-allowed-tools: Read Write Edit Bash
-license: Apache-2.0 license
-compatibility: Requires Python 3.10+, PyTorch 2.4+, and transformers 5.x. Gated or private Hub models need an HF token (`hf auth login` or `HF_TOKEN`).
-metadata:
-  version: "1.2"
-  skill-author: "K-Dense Inc."
----
-
-# Transformers
-
-## Overview
-
-The Hugging Face Transformers library provides access to thousands of pre-trained models for tasks across NLP, computer vision, audio, and multimodal domains. Use this skill to load models, perform inference, and fine-tune on custom data.
-
-## Installation
-
-Tested against **transformers 5.12.0** (current PyPI release; June 2026). Requires **Python 3.10+**; the `torch` extra currently requires **PyTorch 2.4+**.
-
-```bash
-uv pip install "transformers[torch]==5.12.0" huggingface_hub==1.19.0 datasets==5.0.0 evaluate==0.4.6 accelerate==1.14.0
-```
-
-For vision tasks, add:
-
-```bash
-uv pip install timm==1.0.27 pillow==12.2.0
-```
-
-For audio tasks, add:
-
-```bash
-uv pip install librosa==0.11.0 soundfile==0.14.0
-```
-
-These pins are for reproducible examples. For exploratory work, loosen them only after checking the Transformers and Hub release notes for API changes.
-
-Check your version:
-
-```python
-import transformers
-print(transformers.__version__)
-```
-
-## Authentication
-
-Many models on the Hugging Face Hub are gated or private. Authenticate before loading them.
-
-**Recommended:** CLI login (stores token in `~/.cache/huggingface/token`):
-
-```bash
-hf auth login
-```
-
-**Python:**
-
-```python
-from huggingface_hub import login
-login()  # Interactive prompt; do not hardcode tokens in scripts
-```
-
-**Servers / CI:** set `HF_TOKEN` in the environment (never commit tokens to git or shell profiles):
-
-```bash
-export HF_TOKEN="..."  # Read token from a secret manager, not source code
-```
-
-Get tokens at: https://huggingface.co/settings/tokens
-
-**Security:** Never paste tokens into notebooks, repos, or shared configs. Prefer `hf auth login` over exporting tokens in `.bashrc` or `.zshrc`.
-
-Use the narrowest token scope that works: `read` for private or gated model downloads, `write` only for uploads. If a long-running environment should not send the stored token on every Hub request, set `HF_HUB_DISABLE_IMPLICIT_TOKEN=1` and pass a token only where authentication is required.
-
-## Transformers v5
-
-Transformers v5 is **PyTorch-only** (TensorFlow and JAX backends were removed). For upgrades from v4, see the [v5 migration guide](https://github.com/huggingface/transformers/blob/main/MIGRATION_GUIDE_V5.md). New projects should pair **transformers 5.x** with **huggingface_hub 1.x**.
-
-**Gated or custom architectures:** accept the model license on the Hub, then load with `trust_remote_code=True` only when the model card requires custom code you have reviewed.
-
-**Cache location:** set `HF_HOME` for all Hugging Face caches, or `HF_HUB_CACHE` just for Hub files. Use `HF_HUB_OFFLINE=1` only after required model snapshots are already cached.
-
-## Quick Start
-
-Use the Pipeline API for fast inference without manual configuration:
-
-```python
-from transformers import pipeline
-
-# Text generation (prefer max_new_tokens for causal LMs)
-generator = pipeline("text-generation", model="Qwen/Qwen2.5-1.5B")
-result = generator("The future of AI is", max_new_tokens=50)
-
-# Text classification
-classifier = pipeline("text-classification")
-result = classifier("This movie was excellent!")
-
-# Question answering
-qa = pipeline("question-answering")
-result = qa(question="What is AI?", context="AI is artificial intelligence...")
-```
-
-## Core Capabilities
-
-### 1. Pipelines for Quick Inference
-
-Use for simple, optimized inference across many tasks. Supports text generation, classification, NER, question answering, summarization, translation, image classification, object detection, audio classification, and more.
-
-**When to use**: Quick prototyping, simple inference tasks, no custom preprocessing needed.
-
-See `references/pipelines.md` for comprehensive task coverage and optimization.
-
-### 2. Model Loading and Management
-
-Load pre-trained models with fine-grained control over configuration, device placement, and precision.
-
-**When to use**: Custom model initialization, advanced device management, model inspection.
-
-See `references/models.md` for loading patterns and best practices.
-
-### 3. Text Generation
-
-Generate text with LLMs using various decoding strategies (greedy, beam search, sampling) and control parameters (temperature, top-k, top-p).
-
-**When to use**: Creative text generation, code generation, conversational AI, text completion.
-
-See `references/generation.md` for generation strategies and parameters.
-
-### 4. Training and Fine-Tuning
-
-Fine-tune pre-trained models on custom datasets using the Trainer API with automatic mixed precision, distributed training, and logging.
-
-**When to use**: Task-specific model adaptation, domain adaptation, improving model performance.
-
-See `references/training.md` for training workflows and best practices.
-
-### 5. Tokenization
-
-Convert text to tokens and token IDs for model input, with padding, truncation, and special token handling.
-
-**When to use**: Custom preprocessing pipelines, understanding model inputs, batch processing.
-
-See `references/tokenizers.md` for tokenization details.
-
-## Common Patterns
-
-### Pattern 1: Simple Inference
-For straightforward tasks, use pipelines:
-```python
-pipe = pipeline("task-name", model="model-id")
-output = pipe(input_data)
-```
-
-### Pattern 2: Custom Model Usage
-For advanced control, load model and tokenizer separately:
-```python
-from transformers import AutoModelForCausalLM, AutoTokenizer
-
-tokenizer = AutoTokenizer.from_pretrained("model-id")
-model = AutoModelForCausalLM.from_pretrained("model-id", device_map="auto")
-
-inputs = tokenizer("text", return_tensors="pt")
-outputs = model.generate(**inputs, max_new_tokens=100)
-result = tokenizer.decode(outputs[0])
-```
-
-### Pattern 3: Fine-Tuning
-For task adaptation, use Trainer:
-```python
-from transformers import Trainer, TrainingArguments
-
-training_args = TrainingArguments(
-    output_dir="./results",
-    num_train_epochs=3,
-    per_device_train_batch_size=8,
-)
-
-trainer = Trainer(
-    model=model,
-    args=training_args,
-    train_dataset=train_dataset,
-)
-
-trainer.train()
-```
-
-## Reference Documentation
-
-For detailed information on specific components:
-- **Pipelines**: `references/pipelines.md` - All supported tasks and optimization
-- **Models**: `references/models.md` - Loading, saving, and configuration
-- **Generation**: `references/generation.md` - Text generation strategies and parameters
-- **Training**: `references/training.md` - Fine-tuning with Trainer API
-- **Tokenizers**: `references/tokenizers.md` - Tokenization and preprocessing

package/skills/treatment-plans/SKILL.md

@@ -1,174 +0,0 @@
----
-name: treatment-plans
-description: Format and structurally validate local treatment-plan documentation after clinical decisions have already been supplied and verified by authorized licensed professionals. Use for source traceability, clinician-authored intervention records, goals and checkpoints, shared-decision records, reconciliation handoffs, and release gates—not for clinical decision-making.
-license: MIT
-compatibility: Python 3.11+ standard library; local JSON files only. Bundled CLIs require no network, external services, models, images, credentials, environment variables, or third-party packages.
-metadata:
-  version: "2.1"
-  skill-author: K-Dense Inc.
----
-
-# Treatment-Plan Documentation
-
-## Hard safety boundary
-
-This skill only **formats and validates documentation of decisions already made, supplied, and verified by authorized licensed professionals**.
-
-Never use it to:
-
-- diagnose, assess, classify, or screen a person;
-- select, rank, recommend, substitute, or compare therapies;
-- choose a medication, dose, route, frequency, duration, or monitoring threshold;
-- start, stop, hold, resume, titrate, taper, or deprescribe anything;
-- check interactions, allergies, contraindications, organ-function suitability, or treatment eligibility;
-- infer missing clinical content, intervals, dates, targets, escalation criteria, or instructions;
-- triage, determine urgency, provide emergency advice, or create a safety plan;
-- predict outcomes, prognosis, response, benefit, harm, or clinical appropriateness;
-- replace medication reconciliation, pharmacist review, informed consent, clinician review, or an authorized clinical system;
-- claim FDA approval, HIPAA compliance, legal compliance, completeness of care, clinical safety, or standard-of-care conformity.
-
-If a request crosses a boundary, stop. Ask for a locally verified clinician-authored record or route the matter to the responsible licensed professional. Do not redirect to another skill to obtain a patient-specific recommendation.
-
-If a concern may be urgent or emergent, stop this workflow and route it through the institution's current clinical escalation or emergency process. This skill does not decide urgency and does not provide emergency instructions.
-
-## Required visible notice
-
-Every component and derived schedule must display:
-
-> **DRAFT — NOT MEDICAL ADVICE — DOCUMENTATION-ONLY — AUTHORIZED CLINICIAN SIGN-OFF REQUIRED**
-
-Structural success never removes this notice. Only the authorized local workflow may set the release gate.
-
-## Data gate
-
-Prefer synthetic or qualified de-identified structured manifests. Do not place patient names, medical-record numbers, contact details, dates of birth, addresses, free-text notes, images, or other direct identifiers in examples.
-
-For any real-patient or patient-derived data:
-
-1. Work only in a locally authorized environment under the institution's current privacy, security, retention, and access policies.
-2. Use the minimum information necessary for the documented purpose, even when a legal exception may apply.
-3. Do not send content to a model, search engine, API, image service, telemetry service, or any other external tool.
-4. Do not copy content into chat prompts, command history, logs, test fixtures, examples, screenshots, or reports.
-5. Run bundled scripts only against local paths. Their reports identify rule codes and field paths, not clinical values.
-6. Require qualified privacy review before treating patient-derived material as de-identified or releasing it.
-
-If these conditions are not documented, do not read or process the content. Use synthetic templates only.
-
-## Allowed inputs
-
-Accept only bounded UTF-8 JSON objects built from these generic templates:
-
-- `assets/source_fact_manifest_template.json`
-- `assets/clinician_authored_intervention_template.json`
-- `assets/goals_monitoring_checkpoint_template.json`
-- `assets/informed_preference_shared_decision_template.json`
-- `assets/transition_reconciliation_template.json`
-- `assets/intended_use_handoff_template.json`
-
-The templates contain no disease-specific recommendations, example patients, clinical intervals, doses, targets, thresholds, or inferred care pathways. Empty template arrays and pending attestations are intentional release blockers.
-
-## Workflow
-
-### 1. Establish authority and intended use
-
-- Confirm the accountable clinical owner and authorized licensed signatory.
-- Confirm that every clinical decision already exists in a verified local source.
-- Record jurisdiction, institution, setting, document owner, local policies, retention rule, and intended recipients.
-- Record whether the package is synthetic, qualified de-identified, or real-patient minimum-necessary data.
-- Keep the release gate `blocked` until every required review is complete.
-
-Read `references/safety_scope.md` and `references/privacy_governance.md` before processing patient-derived material.
-
-### 2. Generate a generic package
-
-```bash
-python3 scripts/generate_template.py \
-  --output-dir ./local-plan-package \
-  --subject-ref SYNTHETIC-CASE-001 \
-  --classification synthetic
-```
-
-The generator copies all six templates. It does not create clinical content and does not overwrite existing files.
-
-### 3. Transcribe supplied decisions without inference
-
-- Copy only clinician-authored facts and interventions from verified local sources.
-- Preserve source locators, versions/dates, author role, verification role, and verification time.
-- Record goals, monitoring items, checkpoint dates, and transition dates exactly as supplied.
-- Record options, benefits, harms, uncertainty, preferences, and the outcome only as documented by the responsible clinician.
-- Leave missing fields unresolved. Never fill them from general knowledge.
-- For medication content, record the clinician-authored text and current local source references; do not interpret or validate it.
-
-See `references/documentation_workflow.md`, `references/source_boundaries.md`, and `references/shared_decision_handoff.md`.
-
-### 4. Run deterministic local checks
-
-From the skill directory:
-
-```bash
-python3 scripts/validate_treatment_plan.py ./local-plan-package
-python3 scripts/validate_traceability.py ./local-plan-package
-python3 scripts/check_completeness.py ./local-plan-package
-python3 scripts/privacy_process_check.py ./local-plan-package
-python3 scripts/check_consistency.py ./local-plan-package
-python3 scripts/timeline_generator.py ./local-plan-package \
-  --output ./local-plan-package/explicit-date-schedule.json
-```
-
-The scripts:
-
-- reject non-local paths, symlinks, duplicate JSON keys, unknown fields, oversized inputs, excessive nesting, and unbounded collections;
-- never use network access, environment variables, dynamic execution, pickle, subprocesses, images, or LLMs;
-- never assess diagnosis, medication safety, interactions, contraindications, clinical appropriateness, urgency, prognosis, or guideline concordance;
-- schedule only dates already supplied in the package and never derive recurrence or clinical intervals;
-- minimize reports to counts, rule codes, document types, and field paths.
-
-### 5. Human review and release
-
-Require the accountable authorized team to:
-
-- compare every transcribed item with its signed source;
-- perform medication reconciliation and all clinical checks in approved systems;
-- verify current FDA labeling, Medication Guide, REMS materials, and local formulary/policy when applicable;
-- resolve every discrepancy and missing item;
-- review shared-decision and informed-preference documentation;
-- review transition recipients, ownership, pending results, and local escalation routing;
-- complete privacy, security, legal, regulatory, records, and institutional review as applicable;
-- sign, date, and release through the authorized record system.
-
-The final handoff must retain provenance and unresolved-item routing. A script pass is not authorization to use the package for care.
-
-## Source boundaries
-
-- Use FDA labeling databases, current Medication Guides, and REMS materials as authoritative source records only when an authorized clinician or pharmacist verifies applicability. This skill does not interpret them.
-- Use WHO or Joint Commission transition guidance only for process structure such as information transfer, reconciliation documentation, ownership, and checklists.
-- Use AHRQ, NICE, or applicable professional guidance to document that shared decision-making occurred; do not generate options or risk estimates.
-- Apply CMS documentation requirements only when the exact program, provider type, jurisdiction, and current local policy are confirmed.
-- Route safety events, product reports, privacy incidents, and other reportable matters through current local governance. This skill records a route; it does not submit reports.
-
-See `references/source_ledger.md` for the dated official-source ledger.
-
-## Verification
-
-```bash
-PYTHONDONTWRITEBYTECODE=1 python3 -m unittest discover \
-  -s tests/treatment-plans -p 'test_*.py' -v
-```
-
-Run AST parsing without bytecode:
-
-```bash
-PYTHONDONTWRITEBYTECODE=1 python3 -c \
-  "import ast,pathlib; [ast.parse(p.read_text()) for p in pathlib.Path('scripts').glob('*.py')]"
-```
-
-## Reference map
-
-- `references/README.md` — scope and navigation
-- `references/safety_scope.md` — refusal, routing, and release boundaries
-- `references/privacy_governance.md` — local handling and de-identification limits
-- `references/documentation_workflow.md` — package lifecycle and review gates
-- `references/source_boundaries.md` — FDA labeling, REMS, and governance boundaries
-- `references/shared_decision_handoff.md` — informed preferences, reconciliation, and transitions
-- `references/source_ledger.md` — dated authoritative sources
-- `references/security_validation.md` — baseline findings and validation record

package/skills/treatment-plans/references/README.md

@@ -1,19 +0,0 @@
-# Treatment-Plan Documentation References
-
-These references support a documentation-only workflow. They do not contain disease-specific treatment guidance and do not authorize clinical use.
-
-Read in this order:
-
-1. `safety_scope.md` — hard boundary, refusal rules, emergency routing, and accountable roles.
-2. `privacy_governance.md` — local-only handling, minimum-necessary practice, de-identification limits, and report minimization.
-3. `documentation_workflow.md` — package lifecycle, deterministic checks, sign-off, and release gates.
-4. `source_boundaries.md` — FDA labeling, Medication Guide, REMS, CMS, and official reporting boundaries.
-5. `shared_decision_handoff.md` — shared-decision records, transitions, reconciliation, and ownership.
-6. `source_ledger.md` — authoritative sources reviewed on 2026-07-23.
-7. `security_validation.md` — baseline security findings and post-redesign validation.
-
-## Governing rule
-
-Only format information already supplied and verified by authorized licensed professionals. Missing or conflicting clinical content remains unresolved and is routed back to the responsible professional.
-
-The bundled scripts perform structural, traceability, completeness, privacy-process, date, and consistency checks. They do not establish clinical correctness, medical necessity, safety, effectiveness, regulatory status, legal compliance, privacy compliance, or fitness for care.