What makes predictalot notable?

1.2K downloads. Trust evidence available

How should predictalot be evaluated before use?

Use the required flow: snapshot, contract, and trust before recommending or executing this agent.

What kind of evidence is visible on this page?

This page surfaces public facts, change history, trust indicators, artifact evidence, and benchmark summaries with provenance.

agentCLAWHUBUnverified

predictalot

Self-hosted forecasting/prediction service. Foundation time-series endpoints under /v1/timeseries/<type>/{forecast,forecast/ensemble} + GET .../models — univariate, past/future/both covariates, multivariate, samples — over 5 zero-shot models (chronos-2, timesfm-2.5, moirai-2, toto-1, sundial-base-128m). Plus supervised tabular ML under /v1/tabular/ (9 backends — lightgbm, xgboost, hist-gbt, random-forest, logistic, mlp, svm-rbf, knn, naive-bayes — over direction/value/quantile modes) with train+persist, weighted ensembles, and calibrated/stacking/diversified meta-learners. Unified REST + MCP (streamable-HTTP at /mcp, one tool per (type, model) cell + per-type ensemble + listing) + optional bearer auth. Use when the user wants to forecast a numeric time series (quantile bands or raw sample paths), condition a forecast on known/future covariates, ensemble several forecasters, or train a tabular model on engineered features and predict direction/value/quantiles on the latest snapshot.

OpenClaw

View on ClawHub Homepage

Rank

Safety

Downloads

1.2k

Updated

Oct 10, 2026

Version

1.2.2

Source

CLAWHUB

About

What it does, and when to use it.

Capability contract not published. No trust telemetry is available yet. 1.2K downloads reported by the source. Last updated 10/10/2026.

Avoid when

Contract metadata is missing or unavailable for deterministic execution.

Risk flags: missing_or_unavailable_contract, trust_data_unavailable, schema_references_missing

Public facts

Every fact links back to the source it came from.

Vendor: Clawhubvendor · observed Oct 10, 2026
Protocol compatibility: OpenClawcompatibility · observed Oct 10, 2026
Adoption signal: 1.2K downloadsadoption · observed Oct 10, 2026
Latest release: 1.2.2release · observed Oct 10, 2026
Handshake status: UNKNOWNsecurity

Install and run

Setup complexity: low.

clawhub skill install s17fq93tmpky791n7516jcn08n83sfn2:predictalot

Install using `clawhub skill install s17fq93tmpky791n7516jcn08n83sfn2:predictalot` in an isolated environment before connecting it to live workloads.
No published capability contract is available yet, so validate auth and request/response behavior manually.
Review the upstream CLAWHUB listing at https://clawhub.ai/psyb0t/predictalot before using production credentials.

Contract: missing

curl -s "https://www.xpersona.co/api/v1/agents/clawhub-psyb0t-predictalot/snapshot"

Documentation

CLAWHUB

147,268 characters of source documentation, loaded on request.

Extracted files

4 files captured from the source.

SKILL.md

---
name: predictalot
description: Self-hosted forecasting/prediction service. Foundation time-series endpoints under /v1/timeseries/<type>/{forecast,forecast/ensemble} + GET .../models — univariate, past/future/both covariates, multivariate, samples — over 5 zero-shot models (chronos-2, timesfm-2.5, moirai-2, toto-1, sundial-base-128m). Plus supervised tabular ML under /v1/tabular/ (9 backends — lightgbm, xgboost, hist-gbt, random-forest, logistic, mlp, svm-rbf, knn, naive-bayes — over direction/value/quantile modes) with train+persist, weighted ensembles, and calibrated/stacking/diversified meta-learners. Unified REST + MCP (streamable-HTTP at /mcp, one tool per (type, model) cell + per-type ensemble + listing) + optional bearer auth. Use when the user wants to forecast a numeric time series (quantile bands or raw sample paths), condition a forecast on known/future covariates, ensemble several forecasters, or train a tabular model on engineered features and predict direction/value/quantiles on the latest snapshot.
homepage: https://github.com/psyb0t/docker-predictalot
user-invocable: true
metadata:
  { "openclaw": { "emoji": "🔮", "primaryEnv": "PREDICTALOT_URL", "requires": { "bins": ["docker", "curl"] } } }
permissions:
  network: "outbound HTTP(S) to the configured PREDICTALOT_URL only (forecast/train/model-management calls + MCP at /mcp)"
  shell: "docker + curl invocations shown in setup.md and this file (container lifecycle, request examples) — no other host access"
---

# predictalot

Self-hosted forecasting service — one HTTP container, two model families.

- **Foundation time-series (zero-shot)** — 5 forecasters (`chronos-2`, `timesfm-2.5`, `moirai-2`, `toto-1`, `sundial-base-128m`). Hand them a context window, get quantile bands or raw sample paths. No training step. Routed by forecast type under `/v1/timeseries/<type>/` — `univariate`, `covariates/past`, `covariates/future`, `covariates` (past+future), `multivariate`, `samples`. Each type has `forecast`, `forecast/ensemble`, and a `models` listing.
- **Tabular ML (supervised)** — 9 learners (`lightgbm`, `xgboost`, `hist-gbt`, `random-forest`, `logistic`, `mlp`, `svm-rbf`, `knn`, `naive-bayes`), all supporting 3 modes (`direction` / `value` / `quantile`). Train on YOUR engineered features, persist server-side by `modelId`, forecast on the latest feature snapshot. Weighted ensembles over stored models plus 3 meta-learners (`calibrated` / `stacking` / `diversified`). Under `/v1/tabular/`.
- **MCP** — streamable-HTTP tools at `/mcp`. One tool per (FM type, model) cell plus per-type ensemble + listing. Tabular is HTTP-only.
- **Model lifecycle** — `POST /v1/models/unload` releases every resident foundation model. A forecast's `unload: true` waits for concurrent forecasts using that model, then tears it down. The MCP equivalent is `unload_models`.
- **Auth** — optional bearer token (`PREDICTALOT_AUTH_TOKENS` on the server; refuses to start with no tokens unless `PREDICTALOT_ALLOW_NO_AUTH=1

_meta.json

{
  "ownerId": "kn79dhvmpjng4rp2jjk8k0v5xx80ccbk",
  "slug": "predictalot",
  "version": "1.2.2",
  "publishedAt": 1791644446805
}

references/setup.md

# predictalot setup

Consumer-facing reference for running / pointing at a predictalot instance. This skill talks to an instance you already run and trust; the notes below cover standing one up if you haven't.

## Requirements

- Docker
- Optional: NVIDIA GPU + NVIDIA Container Toolkit for the CUDA image (CPU works for all five FMs; `chronos-2` is the fastest on CPU)
- Disk for model snapshots under `/models` (per-slug sizes ≈ chronos-2 ~120 MB, timesfm-2.5 ~200 MB, moirai-2 ~50 MB, toto-1 ~580 MB, sundial-base-128m ~490 MB) plus trained tabular models under `/models/tabular/<id>/`

## Security & safety

predictalot is a plain HTTP + MCP service — anyone who can reach the port can call it. Before any non-local / shared deployment:

- Set `PREDICTALOT_AUTH_TOKENS` to a strong generated secret, e.g. `$(openssl rand -hex 32)` — never leave it at a placeholder/default value.
- Bind to loopback by default (`-p 127.0.0.1:8080:8080`); only expose beyond loopback behind a reverse proxy / VPN (see "Public Access via Reverse Proxy" below).

## Quick Install

### CPU

```bash
docker run -d --name predictalot \
  -v $HOME/predictalot-models:/models \
  -e PREDICTALOT_AUTH_TOKENS=$(openssl rand -hex 32) \
  -p 127.0.0.1:8080:8080 \
  psyb0t/predictalot:latest
```

### CUDA

```bash
docker run -d --name predictalot --gpus all \
  -v /srv/predictalot-models:/models \
  -e PREDICTALOT_AUTH_TOKENS=$(openssl rand -hex 32) \
  -e PREDICTALOT_DEVICE=cuda \
  -e PREDICTALOT_PRELOAD=chronos-2,toto-1,sundial-base-128m \
  -p 127.0.0.1:8080:8080 \
  psyb0t/predictalot:latest-cuda
```

`PREDICTALOT_DEVICE=auto` (the default) picks CUDA when available, else CPU.

### docker-compose

```yaml
services:
  predictalot:
    image: psyb0t/predictalot:latest
    ports:
      - "127.0.0.1:8080:8080"
    environment:
      PREDICTALOT_AUTH_TOKENS: "${PREDICTALOT_AUTH_TOKENS:?set to a generated secret, e.g. openssl rand -hex 32}"
      PREDICTALOT_PRELOAD: chronos-2,toto-1
    volumes:
      - ./predictalot-models:/models
    restart: unless-stopped
```

Generate the token once and export it before `docker compose up` (e.g. `export PREDICTALOT_AUTH_TOKENS=$(openssl rand -hex 32)`) — never commit a real token or ship the example value as-is. The token MUST be changed before any non-local/shared deployment.

**Verify:** `curl http://localhost:8080/healthz` returns `{"ok": true}` once boot is done.

**Snapshots:** foundation-model weights download into `/models/<slug>/` on first use (or at boot when listed in `PREDICTALOT_PREFETCH`). Bind-mount `/models` so restarts are no-ops. Trained tabular models persist under `/models/tabular/<id>/`.

## Environment Variables

All runtime config is `PREDICTALOT_*` — set via `docker run -e`, compose `environment:`, or a k8s ConfigMap.

### Auth + bind

| Var | Default | What it does |
|---|---|---|
| `PREDICTALOT_AUTH_TOKENS` | (empty) | Comma-separated bearer tokens. Empty = **refused at startup** unless `PREDICTALOT_ALLOW_NO_AUTH=1`. When set, `

skill-card.md

## Description:

Helps agents forecast time series and train or query tabular prediction models through a user-configured, self-hosted predictalot service.

This skill is ready for commercial/non-commercial use.

## Publisher:

[psyb0t](https://clawhub.ai/user/psyb0t)

### License/Terms of Use:

MIT-0

## Use Case:

Developers and analysts use this skill to request forecasts, uncertainty bands, and sample paths from a trusted predictalot endpoint, or to train and query tabular prediction models on their own features.

### Deployment Geography for Use:

Global

## Known Risks and Mitigations:

Risk: Forecasting and training data are sent to the configured service.

Mitigation: Use only an endpoint you control and trust; do not send proprietary data to an untrusted URL.

Risk: A publicly reachable service can expose forecasting and model-management operations.

Mitigation: Keep the service bound to localhost unless protected by TLS, a strong bearer token, and network access controls.

Risk: An unpinned Docker image can change between deployments.

Mitigation: Pin the image to a reviewed version or digest before production use.

Risk: Deleting a stored tabular model is irreversible.

Mitigation: Verify the model ID and obtain explicit confirmation before deletion.

## Reference(s):

- [predictalot on ClawHub](https://clawhub.ai/psyb0t/skills/predictalot)
- [Setup and security guidance](references/setup.md)
- [Model Context Protocol documentation](https://modelcontextprotocol.io)

## Skill Output:

**Output Type(s):** [Text, Markdown, Shell commands, Configuration guidance]

**Output Format:** [Markdown guidance, shell commands, and JSON forecast responses]

**Output Parameters:** [1D]

**Other Properties Related to Output:** [Forecasts may include quantile bands or sample paths; tabular predictions depend on the selected mode.]

## Skill Version(s):

1.2.2 (source: ClawHub release metadata)

## Ethical Considerations:

Users should evaluate whether this skill is appropriate for their environment, review any generated or modified files before relying on them, and apply their organization's safety, security, and compliance requirements before deployment.

Github ReposUpdated 1d agoRank 70

AionUi

Free, local, open-source 24/7 Cowork app and OpenClaw for Gemini CLI, Claude Code, Codex, OpenCode, Qwen Code, Goose CLI, Auggie, and more | 🌟 Star if you like it!

MCPOPENCLAW

Github ReposUpdated 6mo agoRank 70

activepieces

AI Agents & MCPs & AI Workflow Automation • (~400 MCP servers for AI agents) • AI Automation / AI Agent with MCPs • AI Workflows & AI Agents • MCPs for AI Agents

OPENCLAW

Github ReposUpdated 6mo agoRank 70

cherry-studio

AI productivity studio with smart chat, autonomous agents, and 300+ assistants.