DaoyuanLi2816 · DaoyuanLi2816 · Aug 12, 2026 · Aug 12, 2026 · Aug 12, 2026
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -6,15 +6,29 @@ All notable changes to miniVERL are recorded here. The format follows
 
 ## [Unreleased]
 
-### verl v0.8 OPD config compiler
+## [0.8.0] - 2026-08-12
+
+### Single-GPU verl v0.8 OPD runtime
 
 - Added the typed `verl-opd-v0.8-single-gpu-v1` configuration profile and an
-  offline `bridge compile-opd` command. Resolved YAML and repeatable dotted
-  overrides compile into deterministic field-by-field compatibility reports.
+  offline compiler plus weight-free `plan` and executable `run` commands.
+  Resolved YAML and repeatable dotted overrides compile into deterministic
+  field-by-field compatibility reports.
 - Unsupported policy-gradient OPD, task-reward mixtures, KL penalties,
   multi-generation, multi-teacher and distributed dimensions fail closed.
   Engine/resource fields are labelled as local reinterpretations rather than
-  upstream-exact behavior; the config-only command loads no model weights.
+  upstream-exact behavior.
+- Added first-class bounded verl Parquet prompts, padded local-HF rollout,
+  exact response-only selection, current-policy/teacher/cache bindings and the
+  pinned verl `forward_kl_topk` loss with token-mean scalar/metric/gradient
+  conformance.
+- Added a packaged Qwen3-0.6B/1.7B NF4 recipe and standard PEFT export. One RTX
+  4080 runtime-conformance run completed its first update in 12.0224 seconds at
+  3.1758 GiB peak reserved VRAM. No alignment-quality comparison was run.
+- Added OPD v2 import/export. Imports publish a canonical prompt profile without
+  inventing an environment or reward; exports preserve PEFT, teacher identity,
+  Parquet bytes and pure OPD overrides without a reward scaffold. Missing base
+  snapshots and teacher-adapter materialization remain explicit launch blockers.
 
 ## [0.7.1] - 2026-08-11
 
@@ -833,7 +847,9 @@ Same-tokenizer only; one trajectory per forward pass; `swap` unavailable for
 quantized models; only Qwen3 and Qwen2 architectures tested; single-seed GPU
 results. The full list is in `docs/limitations.md`.
 
-[Unreleased]: https://github.com/DaoyuanLi2816/mini-verl/compare/v0.7.1...HEAD
+[Unreleased]: https://github.com/DaoyuanLi2816/mini-verl/compare/v0.8.0...HEAD
+[0.8.0]: https://github.com/DaoyuanLi2816/mini-verl/compare/v0.7.1...v0.8.0
+[0.7.1]: https://github.com/DaoyuanLi2816/mini-verl/compare/v0.7.0...v0.7.1
 [0.7.0]: https://github.com/DaoyuanLi2816/mini-verl/compare/v0.6.3...v0.7.0
 [0.6.3]: https://github.com/DaoyuanLi2816/mini-verl/compare/v0.6.2...v0.6.3
 [0.6.2]: https://github.com/DaoyuanLi2816/mini-verl/compare/v0.6.1...v0.6.2

diff --git a/CITATION.cff b/CITATION.cff
@@ -2,14 +2,17 @@ cff-version: 1.2.0
 title: "miniVERL: Auditable single-GPU alignment and distillation runtime"
 message: "If you use miniVERL in your work, please cite it as below."
 type: software
-version: 0.7.1
-date-released: 2026-08-11
+version: 0.8.0
+date-released: 2026-08-12
 license: Apache-2.0
 repository-code: "https://github.com/DaoyuanLi2816/mini-verl"
 url: "https://github.com/DaoyuanLi2816/mini-verl"
 abstract: >-
-  miniVERL is an auditable single-GPU alignment and distillation runtime with a
-  bounded artifact bridge to one pinned verl profile. Its native core is
+  miniVERL is a local single-GPU runtime for a documented subset of verl v0.8
+  on-policy distillation. It consumes typed verl-shaped configuration and
+  Parquet prompts, executes actor rollout, teacher scoring and actor update in
+  one inspectable process, and exports standard PEFT, Parquet and configuration
+  artifacts through a fail-closed pinned bridge. Its native core also supports
   multi-turn, tool-aware on-policy distillation: a student language
   model samples its own tool-using trajectories against deterministic local
   environments, a teacher scores exactly the states the student visited, and
@@ -28,8 +31,8 @@ abstract: >-
   published with their negative and mixed results intact. A fail-closed
   artifact bridge exchanges configs, LoRA adapters, tokenizers and Parquet
   datasets with one pinned verl profile, verifying what it can and refusing to
-  imply the rest: it does not launch distributed jobs and does not establish
-  algorithmic parity with PPO. miniVERL is designed for one personal CUDA GPU,
+  imply the rest: it does not launch distributed jobs or establish broad
+  algorithmic parity. miniVERL is designed for one personal CUDA GPU,
   automatically selects bf16 or fp16, and requires neither Ray nor a cluster.
   Published performance is measured on one RTX 4080; other GPU models use the
   same code path but remain unmeasured. The v0.7 external-alignment study

diff --git a/PROJECT_STATE.md b/PROJECT_STATE.md
@@ -4,9 +4,9 @@ Living build log for **miniVERL** (`mini-verl` / `miniverl` / CLI `miniverl`).
 A checkbox is not evidence: every completed item names the command that was run
 and what it printed.
 
-Last updated: 2026-08-11.
+Last updated: 2026-08-12.
 
-Canonical release state: stable `v0.7.1` (`830a4ca5d873bce4cdcc7c43a44d827b096e8c0c`), development `0.8.0.dev0`.
+Canonical release state: releasing `v0.8.0`.
 Every public version claim is generated from `release-state.yaml` and gated by
 `python scripts/release_state.py --check`.
 
@@ -84,6 +84,26 @@ from construction start. The standard PEFT adapter load passed. This is a
 systems result only; no alignment endpoint or method comparison ran. The
 machine-readable record is `benchmarks/results/rtx4080-verl-opd-runtime-v1.json`.
 
+## v0.8.0 single-GPU verl OPD pivot — PR E
+
+`import-verl --profile verl-opd-v0.8-single-gpu-v1 --config ...` now publishes
+a transactional, canonical OPD profile plus a field-by-field report; it uses
+the same typed compiler as `plan`/`run`, consumes Parquet paths and needs no
+ToolEnvironment or reward. The legacy environment-profile importer remains
+available under its original profile name.
+
+`export-verl` recognizes a completed compatible OPD run and emits the standard
+student PEFT adapter, exact student/teacher identities, byte-preserved Parquet,
+source config, compiled plan and pure `forward_kl_topk` overrides. It emits no
+reward scaffold. Student/teacher/data loadability, upstream parse/tiny-smoke,
+launchability and distributed execution are independent fields. Missing base
+snapshots and teacher-adapter materialization keep new bundles fail-closed and
+`launchable: false`; no distributed command or execution claim is generated.
+
+The generated `docs/generated/verl-opd-v0.8-compatibility.json` binds 72 source
+fields to the pinned compiler fixture. Upstream parse conformance is recomputed
+against verl `7aed6b23`; it does not imply a model launch or distributed job.
+
 ## v0.7.1 Product correction — RELEASE CANDIDATE
 
 Branch `v0.7.1-product-correction` starts from synchronized main

diff --git a/PYPI.md b/PYPI.md
@@ -1,5 +1,5 @@
 <p align="center">
-  <img src="https://raw.githubusercontent.com/DaoyuanLi2816/mini-verl/main/docs/banner.svg" alt="miniVERL — single-GPU LLM post-training" width="880">
+  <img src="https://raw.githubusercontent.com/DaoyuanLi2816/mini-verl/v0.8.0/docs/banner.svg" alt="miniVERL — single-GPU LLM post-training" width="880">
 </p>
 
 <div align="center">
@@ -8,41 +8,41 @@
 [![Build](https://github.com/DaoyuanLi2816/mini-verl/actions/workflows/build.yml/badge.svg)](https://github.com/DaoyuanLi2816/mini-verl/actions/workflows/build.yml)
 [![PyPI](https://img.shields.io/pypi/v/miniverl.svg)](https://pypi.org/project/miniverl/)
 [![Python](https://img.shields.io/badge/python-3.10%20%7C%203.11%20%7C%203.12%20%7C%203.13-blue)](https://www.python.org)
-[![License](https://img.shields.io/badge/license-Apache--2.0-blue)](https://github.com/DaoyuanLi2816/mini-verl/blob/main/LICENSE)
+[![License](https://img.shields.io/badge/license-Apache--2.0-blue)](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.8.0/LICENSE)
 
 </div>
 
 <p align="center">
   <a href="https://pypi.org/project/miniverl/"><strong>PyPI</strong></a> ·
   <a href="https://daoyuanli2816.github.io/mini-verl/"><strong>Stable docs</strong></a> ·
   <a href="https://daoyuanli2816.github.io/mini-verl/dev/">Development docs</a> ·
-  <a href="https://github.com/DaoyuanLi2816/mini-verl/blob/main/README.zh-CN.md">中文</a>
+  <a href="https://github.com/DaoyuanLi2816/mini-verl/blob/v0.8.0/README.zh-CN.md">中文</a>
 </p>
 
-**miniVERL is a local, inspectable single-GPU alignment and distillation
-runtime.** It runs native SFT, DPO, KD and strict OPD recipes, preserves
-assistant-only loss masks and policy-version provenance, and exchanges standard
-HF/PEFT/Parquet artifacts through a fail-closed bridge to one pinned verl
-profile.
+**Run a documented subset of verl-style on-policy distillation on one consumer
+GPU.** miniVERL accepts a typed verl v0.8 OPD profile and Parquet prompts,
+executes actor rollout → teacher scoring → actor update locally, and exports
+standard PEFT/Parquet/config artifacts for scale-out. Native SFT, DPO, KD and
+tool-agent recipes remain available.
 
-PyPI `v0.7.1` is stable; `main` is development. miniVERL is independent from
+PyPI `v0.8.0` is stable; `main` is development. miniVERL is independent from
 verl. It does not claim arbitrary verl YAML execution, distributed execution,
 or full algorithmic compatibility.
 
-## Install and verify in about a minute
+## Pip-only OPD quickstart
 
 ```bash
 python -m pip install "miniverl[train]"
-miniverl doctor
-miniverl demo --fast --output runs/quickstart
-miniverl inspect runs/quickstart/trajectories.jsonl
-miniverl evidence validate alignment-external-v1
+miniverl data sample --format verl-parquet --out prompts.parquet
+miniverl plan --profile verl-opd-v0.8-single-gpu-v1 --config builtin:qwen3-0.6b-1.7b-opd \
+  --set 'data.train_files=["prompts.parquet"]'
+miniverl run --profile verl-opd-v0.8-single-gpu-v1 --config builtin:qwen3-0.6b-1.7b-opd \
+  --set 'data.train_files=["prompts.parquet"]' --dry-run
 ```
 
-The deterministic demo downloads no model and produces typed trajectories, a
-checksummed teacher cache, manifest and report. The evidence command reads
-self-contained package data; it works from a wheel without a Git checkout.
-For schemas and inspection without the ML stack, install `miniverl` alone.
+The sample, plan and dry run need no Git checkout; planning loads no weights.
+Remove `--dry-run` on one CUDA GPU to execute the pinned Qwen3-0.6B/1.7B NF4
+recipe and produce a loadable PEFT adapter. [Follow the OPD quickstart](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.8.0/docs/opd-quickstart.md).
 
 ## Supported hardware and runtime boundary
 
@@ -51,43 +51,36 @@ device-name agnostic, but fit depends on model pair, context, kernels and VRAM.
 Install the matching CUDA-enabled PyTorch build first, then
 `miniverl[train,cuda]`; that extra does not select a CUDA PyTorch wheel.
 Ray, FSDP, Megatron, PPO, GRPO and distributed launch are outside the runtime.
-See the [single-GPU guide](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/single-gpu-guide.md).
+See the [single-GPU guide](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.8.0/docs/single-gpu-guide.md).
 
 ## verl compatibility summary
 
-The bridge targets official verl `v0.8.0` at commit `7aed6b23`. Its verified
-boundary is checksummed standard artifacts plus pinned config-parse and
-model/data-load smoke—not native checkpoint parity or a completed verl job.
-Imports fail closed when dataset, environment, teacher, objective or schedule
-semantics are unresolved; they never substitute calculator tasks or invent an
-unqualified teacher.
+The executable profile targets official verl `v0.8.0` at commit `7aed6b23` and
+supports one actor, one teacher, `n=1`, pure GKD `forward_kl_topk`, token-mean
+aggregation, LoRA/QLoRA and no reward/KL penalty. PG OPD, task-reward mixtures,
+multi-teacher, multimodal and distributed fields fail closed.
 
-Current exports remain `launchable: false`: the base snapshot is absent, the
-reward scaffold fails closed and required mappings remain placeholders. The
-entry point is `launch.template.sh`; readiness, parse/load evidence,
-launchability, distributed execution and semantic parity are separate facts.
-[Read the bridge contract](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/verl-bridge.md).
+Compatible OPD exports contain no reward scaffold. They preserve student and
+teacher identities, Parquet bytes and OPD overrides, but remain
+`launchable: false` until exact base snapshots are materialized. Parse status,
+artifact loadability, launchability and distributed execution are separate.
+[Read the bridge contract](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.8.0/docs/verl-bridge.md).
 
-## One measured systems result
+## Measured RTX 4080 runtime
 
-On one RTX 4080 with Qwen3-0.6B and eight fixed SQLite trajectories, physical
-batch 4 increased dual-model update throughput from 2.369 to 3.866
-trajectories/s. Shared-backbone batch 4 used 2.227 GiB peak reserved memory
-versus 3.035 GiB for dual model while running 10.1% slower. All 12
-preregistered equivalence comparisons passed. This is one workload on one
-machine, not a promise for other GPUs.
-
-![Measured throughput and reserved VRAM for dual-model and shared-backbone runtime cells](https://raw.githubusercontent.com/DaoyuanLi2816/mini-verl/main/docs/consumer-runtime-v1-pareto.svg)
-
-[Consumer Runtime v1 methods and caveats](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/consumer-runtime-v1.md)
+The packaged Qwen3-0.6B/1.7B recipe completed two 16-token rollouts and one OPD
+update with **3.1758 GiB peak reserved VRAM**; the first update completed in
+**12.0224 s** and the standard PEFT adapter reloaded successfully. This proves
+one runtime/artifact path only—no alignment-quality endpoint or method
+comparison ran. [Exact recipe, timings and hashes](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.8.0/docs/opd-quickstart.md).
 
 ## Three paths
 
 | Path | Start with | Concrete artifact | Next |
 | --- | --- | --- | --- |
-| **Align** — use SFT, DPO, KD or OPD only when pilot evidence supports the cost | `miniverl pilot recipes/alignment_policy_conditioned_qwen.yaml` | `alignment-card.json` | [Alignment Lab](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/alignment-lab/alignment-lab-v1.md) |
-| **Distill locally** — strict OPD, shared backbones and padded updates on one CUDA GPU | `miniverl train recipes/qwen_consumer_gpu_shared.yaml --dry-run` | resolved config and revision-pinned PEFT adapter | [Bring your own GPU](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/single-gpu-guide.md) |
-| **Scale out** — convert Parquet, export standard artifacts and inspect the unsupported boundary | `miniverl bridge doctor scaleout-bundle` | `provenance/compatibility-report.json` | [Verified artifact bridge](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/verl-bridge.md) |
+| **Run OPD locally** | `miniverl plan --profile verl-opd-v0.8-single-gpu-v1 --config verl-opd.yaml` | compiled plan, trajectories, targets and PEFT adapter | [Plan and run](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.8.0/docs/opd-quickstart.md) |
+| **Bring a verl config** | `miniverl import-verl --profile verl-opd-v0.8-single-gpu-v1 --config verl-opd.yaml --out local-opd.yaml` | field report plus round-trippable profile | [Compatibility](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.8.0/docs/compatibility.md) |
+| **Move data and artifacts** | `miniverl export-verl --run runs/my-opd --target-verl v0.8.0 --out scaleout` | Parquet + PEFT + OPD override bundle | [Bridge contract](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.8.0/docs/verl-bridge.md) |
 
 ## Research notes and preserved negative evidence
 
@@ -108,7 +101,7 @@ miniverl pilot --builtin-study alignment-external-v1 --json
 The result is `do_not_continue_this_study` and `insufficient_evidence`, not a
 recommendation among SFT/DPO/KD/OPD. Granite Guardian values are unqualified
 selection diagnostics; Granite, PairRM and teacher qualification and the
-reserved final test did not run. [Study and limitations](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/alignment-external/alignment-external-v1.md).
+reserved final test did not run. [Study and limitations](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.8.0/docs/alignment-external/alignment-external-v1.md).
 
 ### Earlier measured alignment case study
 
@@ -118,17 +111,17 @@ ceiling; continued SFT and both OPD variants retained measured regressions.
 The two sandbox safety checks tied at zero while utility still regressed.
 IFEval, XSTest, HarmBench and RewardBench were not executed, and “preference
 win rate” is a deterministic Minipolicy paired outcome, not human preference.
-[Seed-level evidence](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/alignment-lab/alignment-lab-v1.md).
+[Seed-level evidence](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.8.0/docs/alignment-lab/alignment-lab-v1.md).
 
-- [RecoveryBench v1](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/recoverybench/recoverybench-v1.md): frozen-student KD
+- [RecoveryBench v1](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.8.0/docs/recoverybench/recoverybench-v1.md): frozen-student KD
   beat slower fresh-state OPD on the preregistered primary view; the verifier
   gate remained `insufficient_evidence`.
-- [Calculator benchmark](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/benchmarking.md): both negative controls completed
+- [Calculator benchmark](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.8.0/docs/benchmarking.md): both negative controls completed
   normally at 0%; the ambiguous historical protocol-v1 prompt prevents
   attributing failure solely to intrinsic teacher behavior.
-- [Limitations](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/limitations.md), [math](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/math.md),
-  [reproducibility](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/reproducibility.md) and
-  [compatibility policy](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/compatibility.md).
+- [Limitations](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.8.0/docs/limitations.md), [math](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.8.0/docs/math.md),
+  [reproducibility](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.8.0/docs/reproducibility.md) and
+  [compatibility policy](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.8.0/docs/compatibility.md).
 
 New runs establish tokenizer compatibility through structural identity. The
 legacy behavioral fingerprint is only a migration fallback, not identity proof.
@@ -142,8 +135,8 @@ python -m pip install -e ".[dev]"
 pytest -q -m "not gpu and not network"
 ```
 
-Apache-2.0 licensed. See [CONTRIBUTING.md](https://github.com/DaoyuanLi2816/mini-verl/blob/main/CONTRIBUTING.md),
-[SECURITY.md](https://github.com/DaoyuanLi2816/mini-verl/blob/main/SECURITY.md), the [changelog](https://github.com/DaoyuanLi2816/mini-verl/blob/main/CHANGELOG.md) and
-[citation](https://github.com/DaoyuanLi2816/mini-verl/blob/main/CITATION.cff). Project records: [default GPU recipe](https://github.com/DaoyuanLi2816/mini-verl/blob/main/recipes/qwen_consumer_gpu_calc.yaml),
-[frozen calculator result](https://github.com/DaoyuanLi2816/mini-verl/blob/main/benchmarks/results/gpu-calc-hard-equal-update-v2.json)
-and [license](https://github.com/DaoyuanLi2816/mini-verl/blob/main/LICENSE).
+Apache-2.0 licensed. See [CONTRIBUTING.md](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.8.0/CONTRIBUTING.md),
+[SECURITY.md](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.8.0/SECURITY.md), the [changelog](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.8.0/CHANGELOG.md) and
+[citation](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.8.0/CITATION.cff). Project records: [default GPU recipe](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.8.0/recipes/qwen_consumer_gpu_calc.yaml),
+[frozen calculator result](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.8.0/benchmarks/results/gpu-calc-hard-equal-update-v2.json)
+and [license](https://github.com/DaoyuanLi2816/mini-verl/blob/v0.8.0/LICENSE).