Foundry pilot pack

Status: draft for warm-lead conversations, grounded in the approved 2026-05-01 release-candidate pass.

Offer

This pilot is the narrow Foundry V1 proof: a local-only document-search workflow for PDF-heavy businesses running on Apple Silicon with the Hermes adapter path inside Foundry.

What is being shown:

• local ingest -> extract -> index -> search for PDFs
• Foundry CLI setup/health/status/support-bundle path
• Hermes bridge fixed to document-search on api
• local model endpoint fixed to http://127.0.0.1:8085/v1
• signed-PDF preservation through the integrated bridge path

Approved demo command references:

• /Users/lilly/.local/bin/foundry init --node-id rc-demo --adapter hermes --json
• /Users/lilly/.local/bin/foundry doctor --adapter hermes --json
• /Users/lilly/.local/bin/foundry status --adapter hermes --json
• /Users/lilly/.local/bin/foundry support-bundle --adapter hermes
• Full gate: /Users/lilly/clawd/projects/foundry-commercialisation/contracts/run-v1-integration-gate.sh

Target customer

Primary V1 target customer, per locked Forge brief:

• PDF-heavy business
• 200+ PDFs/day
• 50+ fields per document shape
• digitally signed documents matter

Likely buyer/contact hypothesis to validate in discovery (not a proven RC claim):

• technical owner responsible for document operations, privacy, or compliance
• treat exact title, department, and company-fit as discovery questions rather than proven fit

Qualification criteria

Use this pack only if most of the below are true:

1. They have a real document workflow, not a vague “AI strategy” ask.
2. Their pain is around repetitive PDF ingest/search/extraction, especially signed or compliance-sensitive docs.
3. They want local/private processing or have cloud hesitation.
4. They can describe current volume in the rough range of the locked V1 scope.
5. They are open to a bounded pilot, not a fully general automation platform.
6. They accept a human-reviewed operational rollout after the demo; the demo is not autonomous production approval.

What the demo proves

The approved RC evidence proves that, on the reference Mac Studio path:

• Foundry init/doctor/status/support-bundle all run successfully on the Hermes adapter path.
• The bridge defaults are workflow=document-search, entrySurface=api, modelEndpoint=http://127.0.0.1:8085/v1, model=mlx-community/Qwen3-0.6B-8bit.
• Telemetry is off and outbound is disabled.
• The integrated bridge archives a signed PDF byte-for-byte with valid signature/certificate before and after.
• The demo/search proof returns the ingested signed PDF for Acme Marine.

What the demo does NOT prove

Be explicit about these limits:

• It does not prove a broad “AI platform” rollout across many workflows.
• It does not prove unsupervised customer comms, finance actions, or system-of-record writes.
• It does not prove Windows/Linux support.
• It does not prove smaller Mac profiles; the approved brief/contract evidence is still anchored on the Mac Studio M3 Ultra 512GB reference setup.
• It does not prove that raw verifiedRuntimeDocuments in foundry status is a truthful signed-PDF count; the proof source is the integrated bridge artifact, not that counter.
• It does not prove a packaged self-serve installer or customer onboarding flow.

Missing commercial decisions to flag separately

Do not guess these on a call:

• pilot pricing/package choice for this exact warm-lead motion is not frozen in the approved RC materials
• pilot duration/scope boundaries are not frozen in the approved RC materials
• support SLA / response commitment is not frozen in the approved RC materials
• data retention/archive policy for customer pilot documents is not frozen in the approved RC materials

Evidence used

• /Users/lilly/clawd/templates/mc-briefs/forge-project.md
• /Users/lilly/clawd/projects/foundry-commercialisation/contracts/foundry-v1-contract.md
• /Users/lilly/clawd/projects/foundry-commercialisation/release-candidate/README.md
• /Users/lilly/clawd/projects/foundry-commercialisation/release-candidate/CTO-DEMO-SCRIPT.md
• /Users/lilly/clawd/projects/foundry-commercialisation/release-candidate/EXPECTED-OUTPUTS.md
• /Users/lilly/clawd/projects/foundry-commercialisation/release-candidate/EVIDENCE-INDEX.md
• /Users/lilly/clawd/projects/foundry-commercialisation/adapters/hermes/mason-integration-check.md
• /Users/lilly/Documents/Shared/projects/build-foundry-v1-observability-status-pack-for-hermes-bridge/10-cto-summary.md

Use guidance for Ged/Lilly

Lead with the narrow truth: “We can show one local, document-search workflow working end-to-end on a Mac Studio, with signed-PDF preservation and no-cloud posture.”

If the lead wants broader automation, treat that as future-scope discussion, not something already proven.