Enforce generation input budgets before provider spend

[Ankit6149]

Advances #34 with a shared server-enforced input-budget slice.

Implemented:

• reject generation request bodies above 512 KiB before JSON parsing
• validate actual UTF-8 bytes even when Content-Length is absent/untrusted
• stable machine-readable generation_limit.* issues with field, actual, max and safe guidance
• server limits for project name, notes, audience, documentation links, document text, combined context, destinations, output types and canonical source/media record counts
• /api/launch_kit validates all limits before repository/link ingestion and before provider generation
• Studio surfaces the first precise limit diagnostic
• MCP campaign schemas advertise the same field/count ceilings while the API remains the final enforcement boundary
• regression coverage for byte caps, malformed JSON, individual/count limits, combined context, exact boundary acceptance and API/MCP ordering/parity

Still open for #34:

• per-owner/session/IP request rate policy
• concurrent campaign ceilings
• provider-request-count budget
• preflight token/cost estimates and hard spend limits
• stress/abuse acceptance across deployment modes

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated
signal-flow-studio	Ignored	Preview	Sep 25, 2026 11:15am UTC