How baorui-qin/agentforchip works
AgentForChip is a workspace for developing and evaluating an agent-based RTL design and repair system, with a customized DeepSeek Harness and a DeepSeek-compatible inference runtime. The intended user journey is to generate or repair RTL, verify it against trusted evaluation inputs, and retain evidence for assessment. The graph focuses on the sampled P09 RTL tooling and the separately documented inference server/runtime. The tree and excerpts show broad capabilities, but do not establish every runtime connection; dashed, uncited edges mark likely but unverified links.
Main entry points: repair-workflow.ts Repair workflow · snapshots.ts Input snapshots · records.ts Run records · main.py Inference server
Components 16
RTL agent workflow
- Repair workflow
AMD/p09-context-skill-change/rtl-tools/src/phase1/repair-workflow.ts - Model dispatch
AMD/p10-40-trace-change/rtl-tools/src/phase1/p10-model.ts - Prompt context
AMD/p09-real-change/rtl-tools/src/phase1/p09-context.ts - Task catalog
AMD/p09-real-change/rtl-tools/src/phase1/p09-tasks.ts - RTL candidates
AMD/p09-real-change/rtl-tools/src/phase1/candidates.ts - Repair proposals
AMD/p09-context-skill-change/rtl-tools/src/phase1/p09-proposal.ts
Verification evidence
- Input snapshots
AMD/p09-real-change/rtl-tools/src/phase1/snapshots.ts - RTL verification
AMD/p09-real-change/rtl-tools/src/phase1/verification.ts - Repair diagnostics
AMD/p09-real-change/rtl-tools/src/phase1/repair-diagnostics.ts - Run records
AMD/p09-real-change/rtl-tools/src/phase1/records.ts - Experiment comparison
AMD/p09-real-change/rtl-tools/src/phase1/comparison.ts
Inference runtime
- Inference server
AMD/p09-runtime/deepseek-recipe/server-py/main.py - API protocols
AMD/p09-runtime/deepseek-recipe/deepseek-recipe/src/protocol - Stream processing
AMD/p09-runtime/deepseek-recipe/deepseek-recipe/src/stream
Other components
- Engineer
- Inference backend
Connections 17
- Engineer → Repair workflow: starts repair
- Repair workflow → Repair proposals: uses proposals
AMD/p09-context-skill-change/rtl-tools/src/phase1/repair-workflow.ts - Repair workflow → RTL candidates: publishes candidate
AMD/p09-context-skill-change/rtl-tools/src/phase1/repair-workflow.ts - Repair workflow → RTL verification: runs verification
- Model dispatch → RTL candidates: publishes RTL
AMD/p10-40-trace-change/rtl-tools/src/phase1/p10-model.ts - Model dispatch → Prompt context: builds request
- Prompt context → Input snapshots: hashes content
AMD/p09-real-change/rtl-tools/src/phase1/p09-context.ts - Prompt context → RTL candidates: uses interface
AMD/p09-real-change/rtl-tools/src/phase1/p09-context.ts - Task catalog → Input snapshots: hashes evaluation
AMD/p09-real-change/rtl-tools/src/phase1/p09-tasks.ts - Input snapshots → Run records: writes manifest
AMD/p09-real-change/rtl-tools/src/phase1/snapshots.ts - RTL verification → Run records: records runs
- Repair diagnostics → Run records: reads records
AMD/p09-real-change/rtl-tools/src/phase1/repair-diagnostics.ts - Repair diagnostics → Input snapshots: reads snapshots
AMD/p09-real-change/rtl-tools/src/phase1/repair-diagnostics.ts - Experiment comparison → Run records: compares runs
- Inference server → API protocols: converts requests
- Inference server → Stream processing: processes chunks
- Inference server → Inference backend: requests inference
AMD/p09-runtime/deepseek-recipe/server-py/main.py