Repository navigation
chore(release): prepare v1.5.0 and refresh product screenshots - #2395
Conversation
|
Reviewed What I verified (all passed)Version synchronization. Contract digest. The new value Desktop qualification numbers. Screenshots. Resolutions match the SCREENSHOTS.md table byte-for-byte (1600×1084, 1600×1000, 1440×1100). The English Desktop capture shows "This request returned its limit of 10 matches"; the Chinese one shows "本次返回 8 条" — both as documented. Both are anonymous loopback ( Migration documentation. The chains in both guides match Checks. 🟡 P2 — the third copy of the "target release" claim was not synchronized
Both halves of that sentence now contradict Posting this in the summary rather than inline because the file is not part of the diff; there is no changed line to anchor a thread to. Suggested replacement: MergeContent is accurate and the release metadata is internally consistent, so nothing here blocks the merge on its own: mergeable after the one-line README fix (or now, if you would rather defer that to a follow-up). The gateway Two things I could not verify and did not assume: the exact recipe behind the manifest digest |
| 'Request include_receipt during context preparation to inspect selected exact revisions, retrieval and fallback modes, omissions, configuration, and stage timings without returning query or evidence bodies. Receipts are bounded, ephemeral diagnostics; unavailable or oversized receipts leave the original prepared context usable. Normal host injection keeps the default response shape.', | ||
| 'Browse a bounded, paginated Atomic Memory manifest before reading selected exact revisions. New writes retain a short description, with a text preview for explicit writes and descriptions generated within existing extraction calls. Historical descriptions can remain absent; reading the manifest never generates summaries. Extraction instructions also preserve subject, scope, and applicability when facts concern different entities.', | ||
| 'Opt in to final inference request and complete-response checks at owned provider transports. Enforcing mode refuses forbidden content or incomplete inspection; disabled remains the default. Public context admission and short-lived exact Atomic Memory reuse proofs let external executors check their final payloads and current permissions. This protection does not provide general-purpose redaction or a provider cache.', | ||
| 'Introduce the optional Rust PowerContext Gateway in alpha, with OpenAI-compatible Chat Completions and Responses endpoints, request checks before sending, and complete-response checks before release. Opt-in memory injection for Chat Completions and capture of turns that pass release checks support later recall; both are disabled by default. Build and run the Gateway separately, either standalone or alongside Core over a private Unix domain socket. The alpha buffers complete responses for inspection and does not provide real-time token streaming to clients. Its APIs and behavior may change.', |
There was a problem hiding this comment.
[P2] 这一条描述的能力在本分支不可达，而正文没有约束合入顺序
gateway/ 在本 head（53bc5689）和 base（5c249808）都是 0 个文件：
git ls-tree -r --name-only refs/pr/2395 -- gateway | wc -l→0- 全仓
git grep -l "pc-gateway\|powercontext gateway"只命中 4 篇 RFC（docs/{en,zh}/rfcs/2086-*、docs/{en,zh}/rfcs/2199-*），2199 仍是 Proposal - 依赖的 feat(gateway): add Rust PowerContext Gateway executor #2307 当前
state=OPEN，且带未闭合的 P1
正文写的是 "Rust PowerContext Gateway (alpha, #2307)"，只给了编号，没有写"须先合并 #2307"。在 #2307 落地前本分支不可单独发布：站点 changelog 会公开一个拿不到的组件，而仓库里没有任何 Gateway 的构建或运行入口供读者核对 "Build and run the Gateway separately" 这句话。
建议在正文补一条合入顺序约束（例如 Depends on #2307; merge it first, then re-check this bullet），并在 #2307 合并后按新 head 复核本条：届时 gateway/ 应出现在树内，"OpenAI-compatible Chat Completions and Responses endpoints" 与 "buffers complete responses / no real-time token streaming" 才具备可核对的实现。
Resolve the conflict with the v1.5.0 release preparation (oceanbase#2395) and the 27 other commits that landed since the previous merge, including the Rust PowerContext Gateway executor (oceanbase#2307) and historical Server upgrade paths (oceanbase#2394). - `release/upgrade-support.json`: keep the target version and the rationale added upstream, append the sentence about this branch's code-organization changes, and recompute the persistence digest. The generated table and index DDL for all 72 tables is identical to master on the SQLite and MySQL dialects. Follow the moves into code that arrived with the merge: - `evaluation/performance/input_encoding/runner.py`: `_open_runtime` no longer takes a `scheduler_path`, and its callers stop passing one. - Four new input-encoding and context-protection tests stop passing `scheduler_path` to `open_builtin_runtime` and `create_server_app`. - Three new tests import access control from `powercontext.builtin.authz`. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Which issue or RFC does this PR close?
Release preparation for PowerContext v1.5.0; no issue closure.
Merge order: Depends on #2307 (Gateway alpha) and #2394 (historical Server storage normalization). Merge both into
masterbefore merging this release PR or publishing v1.5.0. The release notes describe the final v1.5.0 feature set.Use squash merge for this PR so superseded recording assets from intermediate commits are excluded from the history introduced into
master. The review branch retains its existing commits.origin/master.gateway/is present and its documented endpoints, memory options, and complete-response buffering match the implementation; historical storage normalization is present and its release description remains accurate.schema_review.persistence_sha256with the combined tree, and update the digest and rationale if further persistence changes require it. Runmake upgrade-policy-checkandmake release-check VERSION=1.5.0; publication still requires the existing upgrade acceptance and support-declaration gates.Rationale for this change
Prepare the v1.5.0 installation references, API metadata, release notes, and upgrade guidance together. Replace all Dashboard and Desktop README images with populated recordings of the applications in their built-in dark themes.
README media tradeoff: The six shared GIFs total 21.09 MiB; the English/Japanese README loads 11.47 MiB and the Chinese README loads 9.63 MiB. The 1280-pixel width, 10 fps, and 17–19.5-second loops preserve readable English/Chinese UI text and complete search, Handoff, and exact-revision reading flows at normal speed. This transfer-size budget is intentional for automatically playing previews. Lossless optimization of the English Dashboard Handoff and Desktop samples saved only 2.55% and 3.27%; full-duration 8 fps samples saved 15.89% and 6.77%. The selected assets retain their current readability, timing, and motion.
What changes are included in this PR?
make version-bump VERSION=1.5.0, regenerate Python, Dify, and Desktop contracts, and keep independent plugin/Desktop package versions and historical migration identifiers intact. Align the Server identity regression test with the v1.5 API contract.explain/reasonreplacement, prompt v2 requirements, and the existing upgrade restrictions.v1_5_0_001; correct the release guide to require all twelve installed-release upgrade configurations.112db11e885c38b997da2177e03a4065a7e35dc3, byte-for-byte checks of all 679 runtime files, and the unchanged native real-Server harness. Historical qualification records are preserved.Are there any user-facing changes?
Installation examples and the website changelog target v1.5.0. All Dashboard and Desktop README images are looping dark-theme recordings of populated Memory and Handoff workflows, including native Desktop search and the exact-revision reader.
This preparation PR adds no runtime behavior or database migration.
upgrade-requiredremains in force; the full installed-release upgrade matrix has not been run locally or waived. The registered 1.4.0 source bridge remains outside that matrix. The Desktop recordings are Linux development-preview captures, not Windows/macOS distribution qualification. Creating this PR does not publish a tag, GitHub Release, or package.How was this change tested?
make release-check VERSION=1.5.0pnpm lintandpnpm test— 36 tests passed with the expanded bilingual release notes.make contract-test— 62 passed.pytest tests/test_server.py tests/test_release_version.py tests/test_api_contract.py -q— 143 passed, including the updated Server identity assertion.uv run pytest tests/test_release_version.py tests/test_upgrade_release_gate.py -q— 132 passed, 1 skipped.generate:check,typecheck, andtest— 148 frontend tests passed; Dify generated-contract drift check passed.taskset -c 0,1,8,9 .venv/bin/python desktop/tests/real_server.py— anonymous loopback, Bearer, HTTPS/CA/base path, and injected-provider/enforced-access modes passed against the built wheel. All four Servers exited normally. Earlier unpinned attempts exceeded the cold-start readiness window on CPUs with substantial VM steal; no test assertions or timeouts were changed.uv run prek run -a— all hooks, includingty check, passed.make docs-test— website lint, tests, link checks, and static build, rerun with the GIF assets.AI usage statement
OpenAI Codex (GPT-6), including delegated agents, was used for release preparation, generated-file updates, validation and real application recordings. Demonstration content also used the configured LLM through the real PowerContext runtime.