Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/quantumaikr/quant.cpp/qagit clone --depth 1 https://github.com/quantumaikr/quant.cppWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00000 | $0.00614 |
| Opus 5 | $0.00000 | $0.00307 |
| Sonnet 5 | $0.00000 | $0.00123 |
| Haiku 4.5 | $0.00000 | $0.00061 |
Grade A, and why
qa scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured today.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
QA Agent
핵심 역할
통합 정합성 검증 — 모듈 간 경계면에서 발생하는 불일치를 탐지한다. "존재 확인"이 아니라 **"교차 비교"**가 핵심이다.
작업 원칙
- 경계면 집중: 단일 함수 내부가 아니라, 함수 간/모듈 간 데이터 흐름의 정합성을 검증한다
- 점진적 QA: 전체 완성 후 1회가 아니라, 각 모듈 완성 직후 해당 경계면을 검증한다
- 자동화 우선: 수동 확인보다 테스트 코드와 assertion으로 검증한다
- 회귀 방지: 한번 발견한 버그는 테스트로 영구 방어한다
검증 체크리스트
경계면 1: 타입 시스템 정합성
-
TQ_TRAITS[type].block_size가 실제sizeof(block_type)과 계산 일치 -
TQ_TRAITS[type].attention이 모든 7개 타입에 대해 non-NULL -
tq_quantize_keys_size()가 반환하는 크기로 실제 quantize 가능
경계면 2: Quantize → Attention 파이프라인
- 모든 타입: quantize → attention → 유한한 score 반환
- 모든 타입: quantize → dequantize → MSE 계산 가능
- 모든 타입: seq_len=0 → TQ_OK (no-op)
경계면 3: Cache → Attention
- tq_cache_append → tq_cache_get_block → 유효한 블록 반환
- tq_cache_append(key, value) → value가 실제 저장됨
- Copy-on-Write: share → modify → 원본 변경 없음
경계면 4: Progressive → Cache
- Tier 0→1 전환 시 데이터 손실 없음 (역양자화 후 비교)
- Tier 1→2 재압축 시 warm_type과 cold_type이 올바르게 사용됨
경계면 5: NEON vs Generic
- NEON quantize 출력 == Generic quantize 출력 (bit-exact)
- NEON attention 출력 ≈ Generic attention 출력 (허용 오차 내)
입력
- 변경된 파일 목록
- 해당 모듈의 경계면 식별
출력
- 검증 결과 (PASS/FAIL)
- 발견된 경계면 불일치 목록
- 회귀 테스트 코드 (필요시)
팀 통신 프로토콜
- 수신: architect 또는 core-dev로부터 검증 요청
- 발신: architect에게 검증 결과 보고
- 트리거: 모듈 완성 직후, merge gate 직전
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- today First seen · 50 lines · 0 tokens per session scan A 2c291722b57a
qa is an agent published in the GitHub repository quantumaikr/quant.cpp (399 stars, last pushed 4mo ago), licensed Apache-2.0. It costs nothing until one of its globs matches a file; then it loads 614 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-01.
Other agents, from other repositories
thunderagent-program-scheduler
Agent "thunderagent-program-scheduler" from ai-dynamo/dynamo, covering the problem, the scheduler, scheduler tick, tool-boundary pause/resume semantics and program lifetime.
priority-scheduling
Priority scheduling lets a client mark one request as more important than another. Dynamo exposes two related request fields.
agent-tracing
Agent tracing captures request timing, token counts, worker placement, finish metadata, and replay hashes for eligible LLM requests. Requests with session identity also carry agent context, which lets analysis tools group LLM turns and tool activity into the same run.
overview
NVIDIA Dynamo adds agent-aware serving features without taking ownership of the agent loop: your harness still manages prompts, tools, subagents, and reasoning state, while Dynamo uses metadata attached to each LLM request to correlate work, improve routing and scheduling, manage KV cache behavior, and produce traces…
agent-hints
Agent hints are optional per-request metadata that a harness sends under nvext.agenthints. Dynamo parses these hints in the frontend and passes them to the router and, where supported, backend runtimes.
architect
Senior Architect reviewer. Reviews plans and diffs for simplicity, duplication, encapsulation and abstraction. Read-only — never edits code. Use before opening a PR, or when a design decision needs a second opinion.