PowerContext v1.5.0

PowerContext v1.5.0

Not only memory, but a full story.

English

PowerContext v1.5.0 introduces Source and Artifact plugins, making data ingestion and custom Artifact definition, generation, search, and management extensible. This lays the groundwork for future multimodal data sources and Artifacts tailored to different business needs.

Token efficiency engineering and inference safety are also key improvements: reduce repeated evidence in model inputs, reuse prepared context, and check final requests and complete responses before sending or delivery. This release also adds clearer evidence for context selection, the alpha Rust PowerContext Gateway, improved Desktop and managed local Server workflows, and historical Server storage migration tooling.

Source and Artifact plugins: extend data ingestion and Artifact management

• Add trusted local Source plugins for synchronous writes and continuous collection. Agent hosts can discover plugin capabilities and manage collection targets through MCP; dedicated collection runtimes retain checkpoints and lifecycle state.
• Define and generate custom Artifact types for business workflows, with versioned schemas, exact Source lineage, and declared text or vector search. Included incident-report and ticket-summary plugins demonstrate how project data becomes reusable, searchable results.
• Administrators can load, enable, disable, reload, and unload Artifact plugin bindings. New Artifact bindings start disabled; saved Artifacts remain readable when generation is disabled or plugin code is removed.

Plugin loading requires deployment administration. Local collection process management currently requires Linux. See the Source plugin guide and Artifact plugin examples.

Token efficiency engineering: smaller inputs and less repeated work

• Reduce redundant model input with reversible shared-JSON encoding. Repeated evidence text, metadata, and structure can share one representation while preserving Source identities, order, repeated occurrences, and data types. Encoding covers Atomic Memory extraction/reconciliation and Experience generation/incubation without additional model calls.
• Select encoded input only when it restores the original stage JSON exactly and the complete representation, including decoding instructions, is smaller in both bytes and estimated input tokens. Otherwise, retain the original input. Atomic Memory input budgeting uses the same representation that generation sends.
• Reuse prepared Atomic Memory within a single Scope using short-lived proofs. Valid reuse avoids repeating retrieval, query embedding, and context assembly while rechecking current permissions and evidence. This saves preparation work; the retained context still forms part of the model input.
• Measure the tradeoffs with per-call encoding estimates and an off / shadow / active evaluation workflow covering model quality, actual provider usage, latency, failures, and unknown usage. Local estimates remain separate from provider token accounting.

Encoding is off by default. shadow measures candidates while sending original inputs; experimental active selects eligible smaller inputs. Savings depend on the workload and model, and exact restoration alone does not establish equivalent model understanding or lower total task cost. See input encoding, context reuse, and the encoding evaluation workflow.

Inference safety: check before sending and before delivery

Protection is opt-in and disabled by default. With a supported target configured, enforcing mode provides the following checks; shadow and advisory observe requests without blocking them.

• Inspect final SDK-rendered model requests before transmission and complete responses before delivery at the owned provider transport. Check supported text and tool arguments against configured sensitive values and optional bounded email, private-key, and API-token rules. Admitted content remains unchanged.
• Verify the declared endpoint, model, protocol, context window, and output limit. Unsupported targets, opaque content, or incomplete inspection are refused or withheld in enforcing mode.
• Let external executors check final payloads through the public context admission contract. Exact Atomic Memory reuse revalidates current identity, Scope, permissions, revisions, and dependencies; forgotten memories, revoked access, or expired proofs stop reuse. Executors must honor the admission decision and preserve the admitted payload.
• Keep inference protection in force during input encoding. Check original content, the selected representation, and the complete final request; a policy refusal cannot be bypassed by falling back to unencoded input. Reversible encoding is not privacy redaction.
• Preserve retryable work after mandatory refusals. Refused Memory write-gate or Topic Memory generation work does not publish its result or consume its Source window; saved Sources remain pending for retry.

See inference protection and context admission and reuse.

Understand context selection and inspect memories

• Request include_receipt during context preparation to inspect selected exact revisions, retrieval and fallback modes, omissions, configuration, and stage timings. The diagnostic Receipt contains no query or evidence bodies. If a complete Receipt cannot be returned, the prepared context remains usable.
• Browse a bounded, paginated Atomic Memory manifest before reading selected exact revisions. New memories retain short descriptions; historical descriptions may remain absent, and browsing the manifest does not generate summaries.
• Preserve the subject, scope, and applicability of extracted facts when Sources describe different entities. Automatic host recall keeps its default response shape without requesting a Receipt.

See context preparation and the Receipt API migration notes.

PowerContext Gateway — alpha

• Add an optional Rust Gateway with OpenAI-compatible /gateway/v1/chat/completions and /gateway/v1/responses endpoints. The Gateway delegates request checks before sending and complete-response checks before release to PowerContext Core.
• Opt into Atomic Memory injection for Chat Completions. Capture turns that pass release checks as Sources for background extraction and later recall; capture is available for both endpoint formats. Injection and capture are disabled by default.
• Run the Gateway independently over TCP or alongside Core over a private Unix domain socket. It is built and started separately from the Python Server.

Gateway is an alpha product. It buffers complete responses before release and does not provide real-time token streaming to clients. Its APIs and behavior may change. See the Gateway guide.

Safer Desktop and local Server operations

• Guide connection activation into exact Scope selection and enable actions according to discovered Server compatibility. Recheck Server and Principal identity around private reads and writes; identity changes clear active context, while uncertain writes retain recovery drafts.
• Add bounded session diagnostics and Linux Secret Service credential storage, with an explicit session-only option when persistent storage is unavailable.
• Install and repair fixed managed Server runtimes through independent Ops preflight and apply operations and a constrained Desktop native bridge. Repair preserves configuration and data and leaves the service stopped.
• Check existing managed SQLite storage using private snapshots of the main database and WAL through the exact target runtime.

Desktop remains a preview. Production runtime catalog qualification, signed distribution, and installed-platform acceptance are separate requirements. See Desktop and managed local Server installation.

Agent integrations and more reliable processing

• Add experimental Cindy scoped recall and automatic memory hooks, with a separate native MCP connection for the complete tool catalog.
• Improve pre-compaction lifecycle handling, bound recall queries in Pi, OpenCode, DSH, and Bub, run Codex SessionStart for forked sessions, and support Claude Code hooks on the macOS system Python 3.9.
• Keep all query-embedding waits within one context preparation budget. Reduce repeated reads in Candidate review, Experience incubation, Handoff Reports, and archive compatibility checks; identical rendered code evidence no longer varies with query time.
• Add durable recovery intents for external ownership writes, safe processing failure classifications, and Code publication journals that keep newly orphaned generations reclaimable.

Deployment, database defaults, and documentation

• Deploy API and background roles on Kubernetes with the new Helm chart, using external OceanBase, pinned images, existing Secrets, and explicit database initialization.
• Let Linux and macOS Servers listen on a private Unix domain socket for same-user proxies. Authentication still applies; HTTPS Dashboard access requires explicit external-scheme configuration.
• Default new seekDB and OceanBase selections to the powercontext database while preserving explicit names. Existing seekDB data in test requires selecting that database explicitly; data is not moved automatically. Improve missing-database initialization, seekDB restart recovery, and OceanBase full-text schema reflection.
• Add a bilingual blog with author profiles and populated dark-theme README recordings of Dashboard Memory, Handoff, and native Desktop workflows. Extend evaluation tooling for failed MCP-call recovery, capture ownership, and encoded-input quality.

See the Helm deployment guide. Evaluation fixtures and published measurements do not certify every Agent host, model, or native platform.

Historical Server database compatibility

• Add recognition and normalization of compatible unmanaged historical Server storage on SQLite, embedded seekDB, and OceanBase through server db-migrate. Inspect the actual structure and data state before joining the existing migration chain, preserving Scope identities and compatible extensions.
• Reuse legacy Memory-to-Atomic and Processing conversions, with read-only plans, backup and verification requirements, SQLite rollback, and recoverable native DDL operations.
• Legacy Memory conversion with vector search requires the configured embedding provider during apply; existing Atomic embeddings remain preserved.

Historical normalization does not expand the installed-release acceptance matrix or certify every legacy layout. See the historical storage changes and database migration guide.

Install and upgrade

After the package is published:

uv tool install --force "powercontext[cli,server]==1.5.0"

Retain the extras your deployment uses, such as code or seekdb. Upgrade Server, Python SDK, CLI, MCP, and Agent integrations together, then reopen Agent sessions. Older clients reject the new context capability field even when inference protection is disabled.

Compatibility changes:

• Replace explain with include_receipt in context preparation requests. Replace the removed top-level reason with receipt.retrieval.searches[].fallback_reason, checking that the optional Receipt is present before reading it.
• Update custom Atomic Memory prompts for the v2 extraction and reconciliation definitions.
• Explicitly migrate existing managed databases for owner recovery, Atomic Memory descriptions, Source plugin registrations, and Artifact plugin storage. The target migration revision is v1_5_0_001.

Existing databases require an accepted upgrade path for the exact v1.5.0 artifact. Publication requires all twelve installed-1.3.0 backend/configuration combinations to pass on SQLite, embedded seekDB, and OceanBase, including both seekDB configurations with Code enabled, and the support declaration to be reviewed. Intermediate releases are not added automatically; the registered 1.4.0 bridge remains outside this acceptance matrix.

Drain queued or running 1.3.0 Dream tasks before migration. Stop all writers, back up the complete database and configuration, and review the migration plan before applying it. Preserve the historical Embedding configuration and follow the documented recovery procedure after interruption. Restart services only after full Server verification succeeds. Database migration guide.

Full changelog

---

中文

PowerContext v1.5.0 引入 Source 与 Artifact 插件体系，让数据源接入和自定义 Artifact 的定义、生成、检索与管理都可按需扩展，为后续接入多模态数据源、构建面向不同业务的产物奠定基础。

Token 效率工程与推理安全也是本次升级的重点：减少模型输入中的重复证据，复用已准备的上下文，并在发送前检查最终请求、交付前检查完整响应。此外，本次还提供更清楚的上下文选择依据，推出 alpha 版本的 Rust PowerContext Gateway，完善 Desktop 与受管本地 Server 工作流，并扩展历史 Server 存储迁移工具。

Source 与 Artifact 插件：扩展数据接入与产物管理

• 新增可信本地 Source 插件，支持同步写入与持续采集。Agent Host 可通过 MCP 发现插件能力、管理采集目标；独立采集 Runtime 保留检查点和生命周期状态。
• 通过插件定义并生成面向业务的自定义 Artifact 类型，声明内容版本、精确 Source 血缘，以及文本或向量检索能力。内置事故报告和工单摘要插件，展示如何将项目数据转化为可复用、可检索的结果。
• 管理员可加载、启用、停用、重新加载和卸载 Artifact 插件绑定。新的 Artifact 插件绑定默认停用；生成停用或插件代码移除后，已保存的 Artifact 仍可读取。

加载插件需要部署管理员权限，本地采集进程管理目前要求 Linux。参见 Source 插件指南与 Artifact 插件示例。

Token 效率工程：减少输入 Token 与重复计算

• 通过可逆 shared-JSON 编码减少模型输入中的冗余。重复的证据文本、元数据和结构可以共享一份表示，同时保留 Source 身份、顺序、重复次数与数据类型。覆盖 Atomic Memory 提取／协调、Experience 生成／孵化，不增加额外模型调用。
• 仅在候选能精确还原原始阶段 JSON，且包含解码说明的完整表示在字节数和预估输入 Token 上都更小时，选用编码输入；否则保留原始输入。Atomic Memory 的输入预算与生成时实际发送的表示保持一致。
• 在单个 Scope 内通过短期凭据复用已准备的 Atomic Memory。有效复用可避免重复检索、查询 Embedding 和上下文组装，同时重新核对当前权限与证据。节省的是准备过程的开销，保留的上下文仍会进入模型输入。
• 提供每次调用的编码估算，以及覆盖 off / shadow / active 的评测流程，对照模型质量、供应商实际用量、耗时、失败和未知用量。本地估算与供应商 Token 计量分别记录。

编码默认关闭。shadow 测量候选但仍发送原始输入；试验性 active 选择符合条件的更小输入。节省幅度取决于工作负载和模型，可精确还原本身不代表模型理解等价或任务总成本更低。参见输入编码、上下文复用与编码评测流程。

推理安全：发送前检查，交付前把关

推理保护需显式启用，默认关闭。配置受支持的目标后，enforcing 模式执行以下检查；shadow、advisory 观察请求但不阻断。

• 在自有 Provider 传输边界，检查 SDK 完成格式转换后的最终请求，并在交付前检查完整响应。对支持的文本和工具参数匹配已配置的敏感值，还可启用有限的邮箱、私钥、API Token 格式规则；获准内容保持原样。
• 核对声明的端点、模型、协议、上下文窗口与输出上限。对于不支持的目标、无法检查的内容或不完整的检查，enforcing 模式拒绝发送或交付。
• 外部执行器可通过公开上下文准入契约检查最终载荷。Atomic Memory 精确复用会重新核对当前身份、Scope、权限、版本与依赖；记忆被遗忘、权限撤销或凭据过期时停止复用。执行器需遵守准入决定并保持获准载荷不变。
• 输入编码继续受推理保护约束。检查原始内容、选中的编码表示和完整最终请求，策略拒绝不能通过回退原文绕过。可逆编码不等于隐私脱敏。
• 强制拒绝后保留可重试的工作。Memory 写入 gate 或 Topic Memory 生成遭拒绝时，不发布对应结果、不消费 Source 窗口，已保存的 Source 保持待处理状态，供后续重试。

参见推理保护与上下文准入和复用。

理解上下文选择，查看记忆清单

• 准备上下文时可指定 include_receipt，查看选中的精确版本、检索与回退方式、遗漏原因、配置和阶段耗时。诊断 Receipt 不包含查询或证据正文；无法完整返回回执时，原准备结果仍可使用。
• 通过有界分页的 Atomic Memory 清单筛选记忆，再读取选中的精确版本。新记忆保留简短描述；历史描述可以为空，浏览清单不会触发摘要生成。
• 当 Source 涉及不同实体时，提取结果保留事实对应的主体、范围与适用条件。自动 Host 召回保持默认响应结构，不主动请求 Receipt。

参见上下文准备与 Receipt API 迁移说明。

PowerContext Gateway —— alpha

• 新增可选 Rust Gateway，提供兼容 OpenAI 的 /gateway/v1/chat/completions 与 /gateway/v1/responses 入口。Gateway 委托 PowerContext Core 在发送前检查请求、放行前检查完整响应。
• 可显式启用 Chat Completions 的 Atomic Memory 注入。Chat Completions 和 Responses 均可采集通过放行检查的对话，保存为 Source 供后台提取和后续召回；记忆注入与对话采集默认关闭。
• 支持通过 TCP 独立连接 Core，也可通过私有 Unix domain socket 与 Core 组合部署。Gateway 与 Python Server 分开构建和启动。

Gateway 是 alpha 产品。 当前会缓冲完整响应后再放行，不向客户端提供实时逐 token 输出；其 API 和行为仍可能调整。参见 Gateway 指南。

更可靠的 Desktop 与本地 Server 运维

• 激活连接后引导选择精确 Scope，并按发现的 Server 兼容性启用操作。私有读写前后重新核对 Server 与 Principal 身份；身份变化时清除活跃上下文，结果不确定的写入保留恢复草稿。
• 新增有界会话诊断和 Linux Secret Service 凭据存储；持久存储不可用时，可显式选择仅会话保存。
• 通过独立 Ops 预检和执行操作，以及受限的 Desktop 原生桥接，安装、修复固定版本的受管 Server 运行环境。修复保留配置与数据，完成后服务保持停止。
• 使用数据库主文件和 WAL 的私有快照，由目标版本运行环境检查已有受管 SQLite 存储。

Desktop 仍为预览。生产运行资源目录资格、签名发行与安装平台验收仍需分别完成。参见 Desktop与受管本地 Server 安装。

Agent 集成与更可靠的处理流程

• 新增试验性 Cindy 按 Scope 召回与自动记忆 Hook，通过独立原生 MCP 连接使用完整工具目录。
• 完善压缩前生命周期处理，限制 Pi、OpenCode、DSH 和 Bub 的召回查询长度，在 Codex 分叉会话中执行 SessionStart，并支持在 macOS 系统 Python 3.9 上运行 Claude Code Hook。
• 将一次上下文准备中的所有查询 Embedding 等待限制在同一预算内。减少 Candidate 审核、Experience 孵化、Handoff Report 和归档兼容检查的重复读取；相同代码证据的渲染结果不再随查询时间变化。
• 为外部所有权写入增加持久恢复意图，提供安全的处理失败分类，并通过 Code 发布日志回收新产生的孤立 generation。

部署、数据库默认值与文档

• 新增 Helm chart，使用外部 OceanBase、固定镜像、已有 Secret 和显式数据库初始化，在 Kubernetes 部署 API 与后台角色。
• Linux、macOS Server 可通过私有 Unix domain socket 连接同用户代理。认证要求继续生效；HTTPS Dashboard 访问需显式声明外部协议。
• 新建 seekDB 和 OceanBase 连接默认选择 powercontext 数据库，同时保留显式库名。已有 seekDB 数据位于 test 时需显式选择该库，不自动搬迁数据。完善缺失数据库初始化、seekDB 重启恢复和 OceanBase 全文索引结构识别。
• 新增双语博客、作者资料，以及包含完整内容的暗色 README 录屏，展示 Dashboard Memory、Handoff 和原生 Desktop 工作流。扩展 MCP 调用失败恢复、采集归属和编码输入质量的评测工具。

参见 Helm 部署指南。评测夹具与已发布测量不代表所有 Agent Host、模型或原生平台均已验收。

历史 Server 数据库兼容

• 通过 server db-migrate 识别并归一化 SQLite、嵌入式 seekDB 和 OceanBase 上符合条件的未受管历史 Server 存储。先检查实际结构与数据状态，再接入既有迁移链，保留 Scope 身份与兼容扩展。
• 复用旧 Memory 到 Atomic Memory 及 Processing 转换，保留只读计划、备份与验证要求，并支持 SQLite 回滚和原生 DDL 中断恢复。
• 配置向量检索的旧 Memory 转换在 apply 时需要已配置的 Embedding 服务；已有 Atomic 向量仍保持不变。

历史归一化不扩大真实安装包升级验收矩阵，也不代表所有旧库结构均已通过验收。参见历史存储改动与数据库迁移指南。

安装与升级

包发布后执行：

uv tool install --force "powercontext[cli,server]==1.5.0"

保留当前部署使用的 code、seekdb 等扩展，一起升级 Server、Python SDK、CLI、MCP 与 Agent 集成，并重新打开 Agent 会话。即使推理保护关闭，旧 Client 也会拒绝新的上下文能力字段。

兼容性变化：

• 将上下文准备请求中的 explain 替换为 include_receipt。响应移除顶层 reason；改用 receipt.retrieval.searches[].fallback_reason，读取前先确认可选 Receipt 存在。
• 自定义 Atomic Memory Prompt 需适配 v2 提取与协调定义。
• 为已有受管数据库执行显式迁移，加入所有权恢复、Atomic Memory 描述、Source 插件登记和 Artifact 插件存储。目标迁移修订为 v1_5_0_001。

已有数据库需具备针对准确 v1.5.0 发行物的已验收升级路径。 发布要求真实 1.3.0 到目标产物的十二个后端与配置组合全部通过，覆盖 SQLite、嵌入式 seekDB、OceanBase，包括两个启用 Code 的 seekDB 配置，并完成支持声明审阅。不自动纳入中间发行版；已登记的 1.4.0 桥接仍不在该验收矩阵内。

迁移前排空 1.3.0 中排队或运行中的 Dream 任务。停止全部写入，完整备份数据库与配置，并检查迁移计划。保持历史 Embedding 配置，中断后按文档执行恢复；整体 Server 验证通过后再重启服务。数据库迁移指南。

完整变更

What's Changed

• ci(release): allow explicit publication without upgrade acceptance by @hidb4ai in #2248
• build(deps): bump Swatinem/rust-cache from 2.9.2 to 2.9.3 by @dependabot[bot] in #2238
• build(deps): bump jdx/mise-action from 5.1.1 to 5.2.2 by @dependabot[bot] in #2239
• build(deps): bump pypa/gh-action-pypi-publish from 1.14.2 to 1.14.3 by @dependabot[bot] in #2240
• evaluation(workbuddy): add capture ownership replay and contamination report by @tmt-123 in #2122
• docs(examples): fix restored applicability issue links by @HaoooLee in #2266
• perf(handoff-report): read selection heads in one snapshot by @Inference1 in #2232
• feat(integrations): enforce pre-compaction lifecycle contracts by @beatenevo in #2180
• perf(experience): read the incubation scope-head snapshot once per window by @Inference1 in #2225
• feat(authz): recover external owner writes from durable intents by @AlexStocks in #2187
• feat(ops): install and repair managed local Server runtimes by @knqiufan in #2203
• perf(archive): batch target compatibility checks by @knqiufan in #2204
• ci(desktop):add Ubuntu Linux validation by @CX33071 in #2249
• test(integrations): check the prompt-privacy, bootstrap, and transport copies by @Fengzdadi in #2251
• docs(workbuddy): list every file the manual install needs by @Fengzdadi in #2253
• test(release): isolate parallel gate fixtures from PC_UPGRADE_SOURCE_SHA by @Fengzdadi in #2256
• fix(claude-code): run plugin hooks on the macOS system Python 3.9 by @Fengzdadi in #2254
• build(deps): bump fastmcp from 3.4.9 to 3.4.10 in /evaluation by @dependabot[bot] in #2267
• build(deps): bump fastmcp-slim from 3.4.9 to 3.4.10 in /evaluation by @dependabot[bot] in #2268
• build(deps): bump actions/setup-python from 7.0.0 to 7.1.0 by @dependabot[bot] in #2241
• build(deps): bump astral-sh/setup-uv from 10.0.0 to 10.3.2 by @dependabot[bot] in #2242
• feat: add native Server discovery and fix owner recovery migration by @jackie-cqz in #2212
• feat(memory): enforce subject-bound atomic extraction by @beatenevo in #2078
• fix(context): bound a preparation's query embedding waits by one budget by @Fengzdadi in #2231
• perf(review): batch nested failure evidence reads by @HaoooLee in #2278
• fix(ci): use patched seekdb and require all upgrade configurations by @Zxf-xufeng in #1945
• feat(desktop): collect bounded session observations and safe diagnostics by @sunlight798 in #2234
• feat(evaluation): measure agent recovery from failed MCP calls by @Doar999 in #2146
• ci: shard OceanBase clean migration checks by @zamaoxiaoji in #2243
• feat(desktop): implement Linux credentials and packaged app boundary by @CX33071 in #2297
• feat(website): add bilingual blog and author profiles by @zhanghuidinah in #2295
• feat(cindy): add scoped recall and automatic memory hooks by @PsiACE in #1564
• feat(storage): provision dedicated seekdb and oceanbase databases by @Lkx-JY in #2031
• feat(processing): classify artifact processing failures with safe diagnostics (#2126) by @guamiao in #2189
• test(integrations): characterize the Claude Code and Codex hooks by @Fengzdadi in #2255
• feat(ops): produce pinned managed runtime candidates by @knqiufan in #2288
• feat(desktop): add trusted native Ops bridge by @knqiufan in #2289
• fix(codex): run the SessionStart hook when Codex forks a session by @Fengzdadi in #2290
• fix(integrations): bound the recall query in Pi, OpenCode, DSH, and Bub by @Fengzdadi in #2252
• feat(desktop): guide connection activation into exact Scope selection by @jackie-cqz in #2292
• fix(migrations): accept OceanBase 4.6 FULLTEXT reflection by @Zxf-xufeng in #2294
• feat(recall): add bounded atomic memory manifest by @Tech-Circle-48 in #2192
• feat(source): add source plugin framework and some examples by @Junkrat77 in #2210
• feat: add Helm chart for Kubernetes deployment by @Tsukikage7 in #1854
• fix(code): name a generation before its commit, so a lost directory stays reclaimable by @jasondeng1997 in #2096
• feat(artifact): add managed plugin framework and some example by @Junkrat77 in #2246
• ci: consolidate validation and reduce pull request runner demand by @hidb4ai in #2285
• fix(tests): respect foreign keys during assembly cleanup by @hireview001 in #2346
• fix(tests): reconstruct baseline inference in spawned workers by @hireview001 in #2347
• fix(tests): initialize native Code Server storage before seeding by @hireview001 in #2348
• fix(tests): provision the SeekDB migration fixture database by @hireview001 in #2350
• fix(tests): use dialect-aware provenance index removal by @hireview001 in #2353
• fix(tests): initialize isolated OceanBase assembly storage by @hireview001 in #2355
• fix(tests): preserve configured profile when seeding topics by @hireview001 in #2356
• fix(tests): prepare positive HTTP acceptance databases by @hireview001 in #2357
• fix(tests): clean installed scope projections by @hireview001 in #2360
• fix(tests): audit current memory vectors and scope cleanup by @hireview001 in #2354
• feat(ops): admit existing managed SQLite databases with target Server checks by @knqiufan in #2286
• test(seekdb): enforce native restart and recovery acceptance by @Lkx-JY in #2328
• feat(context): add bounded PreparedContext receipts by @Oxidaner in #2258
• docs(rfc): define context privacy and token efficiency by @hidb4ai in #2086
• docs(rfc): define Rust PowerContext Gateway by @hidb4ai in #2199
• fix(desktop): isolate contexts when Server identity changes by @jackie-cqz in #2291
• fix(code): render identical code evidence without the query time by @jasper-yl in #2366
• fix(eval): do not forbid the auth_mode label in published artifacts by @jasper-yl in #2367
• feat(inference): enforce final request and response privacy checks by @hidb4ai in #2311
• feat(server): support private Unix socket listeners by @hidb4ai in #2312
• fix(eval): trim workspace git history to HEAD and its ancestors by @jasper-yl in #2368
• docs(dsh): correct remote updates and setup diagnostics (#2376) by @knqiufan in #2389
• fix(dsh): exclude current runtime context from recall queries by @YingzuoLiu in #2386
• chore(release): prepare v1.5.0 and refresh product screenshots by @hidb4ai in #2395
• feat(gateway): add Rust PowerContext Gateway executor by @ztelur in #2307
• fix(eval): pin the managed Python for container uv commands by @jasper-yl in #2402
• fix(eval): keep cached and reasoning token counts in SWE-bench Pro reports by @jasper-yl in #2404
• fix(gateway): forward Responses requests intact and replay streams after release by @jasper-yl in #2399
• fix(migrations): support historical Server upgrade paths by @Zxf-xufeng in #2394
• feat(gateway): inject prepared context into Responses requests by @jasper-yl in #2400
• feat(eval): let swebench-pro run take a Codex config and a treatment mode by @jasper-yl in #2409
• feat(eval): show cached and reasoning tokens in the console by @jasper-yl in #2410
• fix(gateway): prepare once per user message within a conversation and replay it by @jasper-yl in #2416
• feat(gateway): inject by default in combined deployments for host conversations by @jasper-yl in #2417
• feat(gateway): recognize native client conversation identities by @jasper-yl in #2419
• feat(gateway): add an operator instruction to Responses requests by @jasper-yl in #2421
• fix(gateway): capture each user turn once with a normalized transcript by @jasper-yl in #2423
• fix(eval): disable Codex web search and flag runs that reach outside the task by @jasper-yl in #2426
• fix(eval): allowlist run network egress by default by @jasper-yl in #2427

New Contributors

• @HaoooLee made their first contribution in #2266
• @zamaoxiaoji made their first contribution in #2243
• @jasper-yl made their first contribution in #2366
• @YingzuoLiu made their first contribution in #2386
• @ztelur made their first contribution in #2307

Full Changelog: powercontext-v1.4.1...powercontext-v1.5.0