-
Notifications
You must be signed in to change notification settings - Fork 0
Use Cases
Use percolation-inversion-compiler-ts when AI agent output should be checked
before it is reused, routed, merged, sent to another agent, or treated as
complete.
Run a compact check before reusing a generated answer:
npx pic-ts agent check --compact --text "Candidate packet: preserve residuals." --profile developmentThis gives a JSON report that shows whether the output is accepted and what remains unresolved.
Use the SDK inside an agent loop:
import { runAgentCheck } from "percolation-inversion-compiler-ts";
const report = runAgentCheck({ agent_output: draftText }, true);Save a JSON report from a build job:
npx pic-ts agent check --compact --text "Candidate packet: preserve residuals." --output agent-check.jsonUpload agent-check.json as an artifact.
Export runtime output as a packet file and inspect it without executing content.
npx pic-ts packet export --report runtime-report.json --output packet.json
npx pic-ts packet inspect --packet packet.jsonAsk for the next finite workflow plan:
npx pic-ts phase plan --compact --text "Candidate packet: preserve residuals." --profile developmentPhase planning is recommendation-only. It does not execute commands.
Emit JSONL records for CCR or another scheduler:
npx pic-ts phase plan --compact --emit ccr-tasks --output tasks.jsonl
npx pic-ts phase gap --compact --emit ccr-residuals --output residuals.jsonlThese files are candidate task and residual records. The downstream runtime must still apply leases, authority checks, provider policy, and settlement rules.
Observe several local reports together:
npx pic-ts phase lab init --output-dir .pic-lab
npx pic-ts phase lab ingest --store .pic-lab --report runtime-report.json
npx pic-ts phase lab graph --store .pic-lab
npx pic-ts phase lab observe --store .pic-labThis helps find accepted packets, candidate-only packets, missing evidence, and blocked execution paths.
Use BIT, SQOT, ALT, TRC, and ecology commands to inspect bottlenecks, queue pressure, abstraction reuse, typed traces, and execution-available paths.
These commands are useful for review and routing. They do not grant execution authority.
Check whether a trace is structured enough for scoped operation review:
npx pic-ts trc trace-normalize --input trace.json --output trace_nf.json
npx pic-ts trc trace-check --trace trace_nf.json --output trace_check.jsonThis can report operation_ready=true when authority, resources, rollback,
witnesses, schedule, and tolerance data are present. It still does not execute
the operation.
Common uses include AI agent output checking, LLM output validation, JSON schema validation, workflow report generation, residual ledger preservation, missing obligation tracking, packet inspection, Phase Ecology Lab observation, effective packet graph generation, Node.js agent runtime integration, Python-compatible PIC reports, CCR task JSONL, CCR residual JSONL, TRC operation readiness, TraceNF, operation_ready, ASI-proxy acceleration, and npm workflow diagnostics.
- Home
- Quickstart
- Getting Started
- Acceleration Measurement
- Operation Guide
- AFST Satisfaction Flux
- Python TypeScript Compatibility
- Security
- Node Only Agent Loop
- ASI Proxy Loop v0.9
- Token Reports and Agent Loop
- ASI Proxy Acceleration
- CCR Interop
- MCP And A2A Safety
- Operation Gate
- Phase Ecology Lab
- CLI Recipes
- JavaScript SDK
- Diagnostic Commands
- How To Read The JSON
- Agent Guide
- Agent Messages and Packets
- Schemas and Conformance
- Safety and Limits
- Compatibility With Python
- npm Package and Release Checks
- Use Cases
- Core Concepts
- Concepts in Plain Language
- What PIC Does Not Do
- Theory Map
- Glossary
- FAQ
- v1.1 Release Notes
- v0.9.0 Release Notes
- v0.8.0 Release Notes
- v0.7.0 Release Notes
- v0.6.0 Release Notes
- v0.5.0 Release Notes