CoolFace
Apppublic

Prithvi1312/aegis-pm

sourceHugging Faceapache-2.0updated 5mo agoView on Hugging Face
1likes
App README

Aegis PM — Hosted PRD Scorer

Paste a PRD, get a 100-point score across 10 dimensions plus rewrite suggestions. Powered by the open-source Aegis PM analyzer.

What this Space does

  • —Runs the rules-only scorer by default — deterministic, no API key needed, $0/call
  • —Optionally runs the LLM scorer if a cloud-LLM key is set as a Space secret
  • —Scores against the same 100-point rubric used in the public eval leaderboard

Optional: enable richer LLM-mode analysis

To unlock the "Use LLM" checkbox in the UI, add a Space secret in this Space's Settings → Variables and secrets:

SecretWhen to set it
GROQ_API_KEYFree tier, no card — get one at https://console.groq.com
OPENROUTER_API_KEYFree models routed — https://openrouter.ai
GEMINI_API_KEYGoogle free tier — https://aistudio.google.com/apikey
OPENAI_API_KEYPaid, gpt-4o-mini ~$0.0003/analysis
ANTHROPIC_API_KEYPaid, claude-sonnet ~$0.003/analysis
AEGIS_REDACT=1Recommended — strips PII from prompts before any cloud call

Without any of these set, the Space runs in rules-only mode and the "Use LLM" checkbox does nothing (graceful — the UI shows a notice).

Deploy your own

  1. 1.Fork github.com/prithvi471/opensoirceor
  2. 2.On Hugging Face, click Create Space → Gradio → from-GitHub-fork
  3. 3.Set this file (HF_SPACE_README.md) as your Space README.md, OR copy its frontmatter to a new README.md in your Space root
  4. 4.Optional: add the secrets above
  5. 5.The Space auto-builds from app.py

See DEPLOY.md in the repo for the full one-page guide.

Privacy

Anything you paste here is sent to the configured LLM provider (if "Use LLM" is checked) — see TRUST.md. The default rules-only path makes no network calls — your PRD never leaves the Space's container.

For finserv / regulated content, run the local Ollama path on your own infra instead of using this hosted Space.