stable-harness 0.0.140 → 0.0.144

package/docs/evaluation/0.1.0-bfcl-targeted-review-matrix.zh.md

@@ -7,7 +7,7 @@
 - 标准数据：Berkeley Function Calling Leaderboard BFCL v4 single-turn categories
 - 模型：`qwen3.5:2b, qwen3.5:4b, qwen3.5:9b, gemma4:e2b`
 - 模式：`raw, repair, review`
-- 远端 Ollama：`https://ollama-rtx-4070.easynet.world`
+- 远端 Ollama：`https://ollama-rtx-4070.easynet.world/v1`
 - 用例数：`1800`
 - 说明：这是 BFCL 标准数据上的 stable-harness guard/repair 测试，不是 BFCL 官方 leaderboard 分数。
 

package/docs/evaluation/0.1.0-bfcl-tool-guard.zh.md

@@ -6,7 +6,7 @@
 
 - 标准数据：Berkeley Function Calling Leaderboard BFCL v4 single-turn categories
 - 模型：`qwen3.5:4b`
-- 远端 Ollama：`https://ollama-rtx-4070.easynet.world`
+- 远端 Ollama：`https://ollama-rtx-4070.easynet.world/v1`
 - 用例数：`1390`
 - Repair：`on`
 - 说明：这是 BFCL 标准数据上的 stable-harness guard/repair 测试，不是 BFCL 官方 leaderboard 分数。

package/docs/granite-tool-calling-comparison.zh.md

@@ -6,7 +6,7 @@
 
 - Default Ollama model id: `qwen3.5:0.8b`
 - Default Ollama options: `num_ctx=2048`, `num_predict=256`
-- Granite baseline model id: `granite4.1:3b`
+- Granite baseline model id: `qwen3.5:4b`
 - Hugging Face model id: `ibm-granite/granite-4.1-3b`
 - Ollama native endpoint: `/api/chat`
 - vLLM/SGLang endpoint: OpenAI-compatible `/v1/chat/completions`
@@ -76,11 +76,11 @@ Parser support varies by backend and model template. Keep parser selection as ba
 
 Validated in this checkout:
 
-- local Ollama `http://127.0.0.1:11434`, model `granite4.1:3b`: 3/3 passed.
-- remote Ollama `https://ollama-rtx-4070.easynet.world`, model `granite4.1:3b`: 3/3 passed.
+- local Ollama `https://ollama-rtx-4070.easynet.world/v1`, model `qwen3.5:4b`: 3/3 passed.
+- remote Ollama `https://ollama-rtx-4070.easynet.world/v1`, model `qwen3.5:4b`: 3/3 passed.
 - local Ollama through OpenAI-compatible `/v1`, `tool_choice=required`: 3/3 passed.
 - local Ollama through OpenAI-compatible `/v1`, `tool_choice=named`: 3/3 passed.
-- CUDA host Ollama `http://192.168.0.201:11434`, model `granite4.1:3b`: 3/3 passed.
+- CUDA host Ollama `https://ollama-rtx-4070.easynet.world/v1`, model `qwen3.5:4b`: 3/3 passed.
 - CUDA host Ollama through OpenAI-compatible `/v1`, `tool_choice=required`: 3/3 passed.
 - CUDA host Ollama through OpenAI-compatible `/v1`, `tool_choice=named`: 3/3 passed.
 - CUDA host vLLM `http://192.168.0.201:8000/v1`, model `ibm-granite/granite-4.1-3b`, `tool_choice=auto`: 3/3 passed.
@@ -172,7 +172,7 @@ Expanded matrix: 15 cases. The additional cases cover closed-ticket search, Hong
 
 Ollama models currently visible on `192.168.0.201` include:
 
-- Small/medium chat candidates: `qwen3.5:0.8b`, `qwen3:0.6b`, `lfm2.5-thinking:latest`, `granite4.1:3b`, `qwen3.5:2b`, `qwen3.5:4b`, `qwen3:latest`, `qwen3.5:9b`, `qwen2.5:7b-instruct`, `gemma4:e2b`, `gemma4:e4b`, `gpt-oss:latest`.
+- Small/medium chat candidates: `qwen3.5:0.8b`, `qwen3:0.6b`, `lfm2.5-thinking:latest`, `qwen3.5:4b`, `qwen3.5:2b`, `qwen3.5:4b`, `qwen3.5:4b`, `qwen3.5:9b`, `qwen3.5:4b`, `gemma4:e2b`, `gemma4:e4b`, `gpt-oss:latest`.
 - Very large or slower candidates not included in the fast matrix: `qwen3.6:27b`, `qwen3.6:35b`, `gemma4:26b`, `qwen3.5:27b`, `nemotron-cascade-2:30b`, `lfm2:latest`.
 - Non-chat/specialized models not included: `nomic-embed-text:latest`, `glm-ocr:q8_0`.
 
@@ -184,9 +184,9 @@ Fast matrix results:
 | Ollama | `qwen3.5:4b` | native auto tools | 15/15 | none |
 | Ollama | `qwen3.5:9b` | native auto tools | 15/15 | none |
 | Ollama | `qwen3.5:0.8b` | native auto tools | 14/15 | `freshness` day instead of month |
-| Ollama | `granite4.1:3b` | native auto tools | 14/15 | Chinese news query became unnatural text |
-| Ollama | `qwen3:latest` | native auto tools | 14/15 | ExampleCo ticker became `EXCO` |
-| Ollama | `qwen2.5:7b-instruct` | native auto tools | 14/15 | path with space was collapsed |
+| Ollama | `qwen3.5:4b` | native auto tools | 14/15 | Chinese news query became unnatural text |
+| Ollama | `qwen3.5:4b` | native auto tools | 14/15 | ExampleCo ticker became `EXCO` |
+| Ollama | `qwen3.5:4b` | native auto tools | 14/15 | path with space was collapsed |
 | Ollama | `gemma4:e2b` | native auto tools | 14/15 | missed Chinese ExampleCo stock tool call |
 | Ollama | `gemma4:e4b` | native auto tools | 14/15 | missed Chinese ExampleCo stock tool call |
 | Ollama | `lfm2.5-thinking:latest` | native auto tools | 13/15 | namespace typo and HK market error |

package/docs/guides/getting-started.md

@@ -56,8 +56,8 @@ For a local or private OpenAI-compatible endpoint, keep the same model catalog
 shape and change the environment variables:
 
 ```bash
-export STABLE_HARNESS_MODEL=granite4.1:3b
-export STABLE_HARNESS_OPENAI_BASE_URL=http://127.0.0.1:11434/v1
+export STABLE_HARNESS_MODEL=qwen3.5:4b
+export STABLE_HARNESS_OPENAI_BASE_URL=https://ollama-rtx-4070.easynet.world/v1
 export OPENAI_API_KEY=local
 ```
 
@@ -87,21 +87,29 @@ Print runtime trace lines:
 stable-harness -w . --agent orchestra --trace "Call the echo tool with hello."
 ```
 
-## Start The OpenAI-Compatible Facade
+## Start The Runtime Daemon
 
 ```bash
-stable-harness start -w . --port 8642
+stable-harness start -w .
+stable-harness serve -w .
 ```
 
-Clients can point at:
+Clients can point at the fixed local gateway paths:
 
 ```text
-http://127.0.0.1:8642/v1
+stable-harness-studio: https://shs.easynet.world/?baseUrl=http%3A%2F%2F127.0.0.1%3A56789
+LangSmith Studio: https://smith.langchain.com/studio/?baseUrl=http%3A%2F%2F127.0.0.1%3A56789%2Fprotocols%2Flanggraph
+Runtime API: http://127.0.0.1:56789/runtime/v1
+OpenAI-compatible: http://127.0.0.1:56789/protocols/openai/v1
+LangGraph-compatible: http://127.0.0.1:56789/protocols/langgraph
 ```
 
 The facade is a protocol adapter around the same Stable Harness runtime. It is
 not a separate execution path.
 
+The hosted Stable Harness Studio also accepts an explicit gateway parameter:
+`https://shs.easynet.world/?baseUrl=<stable-harness-gateway-url>`.
+
 ## Develop The Framework
 
 Clone the repository only when you are changing Stable Harness itself:

package/docs/guides/index.md

@@ -32,8 +32,9 @@ embed it, operate it, or explain why it exists.
   relative to agent frameworks, protocol gateways, and orchestration systems.
 - [Product boundary](../product-boundary.md): the guardrails that keep the project
   framework-generic and passthrough-first.
-- [Compatibility matrix](../compatibility-matrix.md): current backend and protocol
-  surfaces.
+- [Protocol feature coverage matrix](../protocols/coverage-matrix.md): side-by-side
+  runtime feature and API support across the Runtime API, OpenAI-compatible,
+  LangGraph-compatible, ACP, A2A, and AG-UI surfaces.
 
 ## Engineering References
 
@@ -43,7 +44,8 @@ embed it, operate it, or explain why it exists.
 - [OpenAI-compatible protocol](../protocols/openai-compatible.md)
 - [LangGraph-compatible protocol](../protocols/langgraph-compatible.md)
 - [HTTP runtime protocol](../protocols/http-runtime.md)
-- [ACP, A2A, and AG-UI protocol facades](../protocols/agent-protocols.md)
+- [ACP, A2A, and AG-UI protocol facades](../protocols/protocol-facades.md)
+- [Protocol feature coverage matrix](../protocols/coverage-matrix.md)
 
 The short version: keep execution semantics upstream-native, define product
 inventory in YAML, and use Stable Harness for lifecycle, governance,

package/docs/guides/integration-guide.md

@@ -23,7 +23,7 @@ stable-harness workflow render review-shell -w ./workspace
 stable-harness workflow inspect review-shell -w ./workspace
 ```
 
-The CLI is intentionally thin. When the Stable Runtime daemon for the workspace
+The CLI is intentionally thin. When the native runtime daemon for the workspace
 is already running, request commands submit work through HTTP and consume
 runtime updates through SSE. When no daemon is available, the CLI falls back to
 an in-process runtime for local development.
@@ -32,7 +32,7 @@ Clients can choose the connection mode explicitly:
 
 ```bash
 stable-harness -w ./workspace --runtime auto "Review the release evidence."
-stable-harness -w ./workspace --runtime daemon --daemon-url http://127.0.0.1:8641 "Review the release evidence."
+stable-harness -w ./workspace --runtime daemon --daemon-url http://127.0.0.1:56789/runtime/v1 "Review the release evidence."
 stable-harness -w ./workspace --runtime local "Review the release evidence."
 stable-harness console -w ./workspace --runtime daemon
 ```
@@ -40,30 +40,31 @@ stable-harness console -w ./workspace --runtime daemon
 | Mode | Behavior |
 | --- | --- |
 | `auto` | Use the matching workspace daemon when available; otherwise run in-process. |
-| `daemon` | Require a matching Stable Runtime daemon and fail clearly if unavailable. |
+| `daemon` | Require a matching native runtime daemon and fail clearly if unavailable. |
 | `local` | Always create an in-process runtime for this CLI process. |
 
 Clients can also choose the wire protocol used for daemon/server requests:
 
 ```bash
-stable-harness -w ./workspace --protocol stable-runtime "Review the release evidence."
-stable-harness -w ./workspace --protocol a2a --protocol-url http://127.0.0.1:8650 "Review the release evidence."
-stable-harness -w ./workspace --protocol agui --protocol-url http://127.0.0.1:8650 "Review the release evidence."
-stable-harness -w ./workspace --protocol acp --protocol-url http://127.0.0.1:8650 "Review the release evidence."
-stable-harness console -w ./workspace --protocol a2a --protocol-url http://127.0.0.1:8650
+stable-harness -w ./workspace --protocol runtime "Review the release evidence."
+stable-harness -w ./workspace --protocol a2a --protocol-url http://127.0.0.1:56789/protocols "Review the release evidence."
+stable-harness -w ./workspace --protocol agui --protocol-url http://127.0.0.1:56789/protocols "Review the release evidence."
+stable-harness -w ./workspace --protocol acp --protocol-url http://127.0.0.1:56789/protocols "Review the release evidence."
+stable-harness console -w ./workspace --protocol a2a --protocol-url http://127.0.0.1:56789/protocols
 ```
 
 | Protocol | Behavior |
 | --- | --- |
-| `stable-runtime` | Native Stable Runtime HTTP + SSE with request/session/trace/operator features. |
+| `runtime` | Native runtime HTTP + SSE with request/session/trace/operator features. |
 | `a2a` | Sends request turns through the A2A facade. |
 | `agui` | Sends request turns through the AG-UI HTTP+SSE facade. |
 | `acp` | Sends request turns through the ACP HTTP JSON-RPC custom transport. |
 
-`--protocol-url` overrides the configured agent protocol server. A2A, AG-UI,
-and ACP carry explicit Stable Runtime request extensions for CLI parity, but
-they do not replace the Stable Runtime operator API. Use `stable-runtime` for
-traces, approvals, memory administration, and full session inspection.
+`--protocol-url` overrides the configured protocol facade server. A2A, AG-UI,
+and ACP carry explicit native runtime request extensions for CLI parity, but
+the gateway also exposes the native runtime control
+plane at `/runtime/v1/*` for traces, approvals, memory administration,
+artifacts, workflows, and full session inspection.
 
 The CLI prints the selected runtime path to stderr:
 
@@ -91,10 +92,10 @@ supports ordinary prompts plus slash commands:
 | `/exit` | Leave console mode. |
 
 When the console uses `--protocol a2a`, `--protocol agui`, or `--protocol acp`,
-the selected protocol carries the current session ID for prompt turns, but the
-full operator commands still belong to `stable-runtime`. Use
-`--protocol stable-runtime` when the console must list sessions, inspect
-requests, manage memory, clear sessions, or stream native control-plane events.
+the selected protocol carries prompt turns while slash-command operator features
+use the same gateway's `/runtime/v1/*` route. This keeps `/sessions`,
+`/requests`, `/memory`, `/clear`, `/debug`, and native event streaming available
+when switching protocol.
 
 ## Embedded Runtime
 
@@ -146,31 +147,30 @@ is the native `StableHarnessRuntime` object returned by
 Start the server:
 
 ```bash
-stable-harness start -w ./workspace --host 127.0.0.1 --port 8642
+stable-harness start -w ./workspace --host 127.0.0.1
 ```
 
-`stable-harness start` exposes the first-party Stable Runtime control plane,
-the OpenAI-compatible facade, the LangGraph-compatible server, and optional
-ACP/A2A/AG-UI facades when enabled by runtime config. Point first-party clients
-at:
+`stable-harness start` exposes the first-party native runtime control plane,
+the OpenAI-compatible facade, the LangGraph-compatible server, and ACP/A2A/AG-UI
+facades. Point first-party clients at:
 
 ```text
-http://127.0.0.1:8641
+http://127.0.0.1:56789/runtime/v1
 ```
 
 Point OpenAI-compatible clients at:
 
 ```text
-http://127.0.0.1:8642/v1
+http://127.0.0.1:56789/protocols/openai/v1
 ```
 
 Use this path when an existing product or evaluation harness already speaks
 OpenAI-compatible chat completions. Stable Harness still owns workspace loading,
 runtime events, tool-gateway policy, memory lifecycle, and adapter selection.
 
-## Stable Runtime HTTP + SSE
+## native runtime HTTP + SSE
 
-Use Stable Runtime HTTP + SSE when the caller needs Stable Harness concepts
+Use native runtime HTTP + SSE when the caller needs Stable Harness concepts
 directly: request IDs, sessions, traces, spans, approvals, cancellation,
 artifacts, memory lifecycle, runtime inspection, and real-time control-plane
 updates.
@@ -180,16 +180,16 @@ compatible `/v1`, LangGraph-compatible, ACP, A2A, and AG-UI servers are
 ecosystem facades that map to this protocol instead of adding new runtime
 semantics.
 
-See [Stable Runtime HTTP + SSE](../protocols/http-runtime.md).
+See [native runtime HTTP + SSE](../protocols/http-runtime.md).
 
 ## Client Connection Choices
 
 | Client kind | Recommended connection | When to use it |
 | --- | --- | --- |
 | CLI one-shot | `--runtime ...` plus `--protocol ...` | Use runtime mode for process selection and protocol mode for wire compatibility. |
-| CLI console | Same runtime and protocol options as one-shot CLI | Use `stable-runtime` for full slash-command operator features; use A2A/AG-UI/ACP for protocol turn testing. |
+| CLI console | Same runtime and protocol options as one-shot CLI | Use `runtime` for full slash-command operator features; use A2A/AG-UI/ACP for protocol turn testing. |
 | Embedded app | SDK / in-process runtime | Use when the application owns the process lifecycle. |
-| First-party UI | Stable Runtime HTTP + SSE | Use for sessions, request inspection, approvals, artifacts, memory, traces, and cross-client realtime events. |
+| First-party UI | native runtime HTTP + SSE | Use for sessions, request inspection, approvals, artifacts, memory, traces, and cross-client realtime events. |
 | OpenAI clients | OpenAI-compatible `/v1` | Use for existing chat-completion clients and evaluation harnesses. |
 | LangGraph Studio | LangGraph-compatible server | Use for LangGraph/LangSmith Studio workflows. |
 | A2A clients | A2A facade | Use when the caller already speaks A2A task/message APIs. |
@@ -200,36 +200,37 @@ Session support is intentionally layered:
 
 | Surface | Session level |
 | --- | --- |
-| Stable Runtime HTTP + SSE | Full operator session API: request history, events, deletion, memory, approvals, and inspection. |
+| native runtime HTTP + SSE | Full operator session API: request history, events, deletion, memory, approvals, and inspection. |
 | SDK / in-process runtime | Full operator session API through `StableHarnessRuntime`. |
 | OpenAI-compatible `/v1` | Chat continuity through session headers or metadata. |
-| LangGraph-compatible | Maps LangGraph `thread_id` to Stable Runtime `sessionId`. |
-| A2A | Maps A2A `contextId` or `taskId` to Stable Runtime `sessionId`. |
-| AG-UI | Maps AG-UI `threadId` or `sessionId` to Stable Runtime `sessionId`. |
-| ACP HTTP JSON-RPC | Uses ACP `session/*` methods and maps `sessionId` to Stable Runtime. |
+| LangGraph-compatible | Maps LangGraph `thread_id` to native runtime `sessionId`. |
+| A2A | Maps A2A `contextId` or `taskId` to native runtime `sessionId`. |
+| AG-UI | Maps AG-UI `threadId` or `sessionId` to native runtime `sessionId`. |
+| ACP HTTP JSON-RPC | Uses ACP `session/*` methods and maps `sessionId` to native runtime. |
 
 Protocol facades preserve continuity for their ecosystems, but they do not
-replace the Stable Runtime operator API.
+replace the native runtime operator API.
 
 ## ACP, A2A, and AG-UI Facades
 
-Enable the combined agent protocol server when an external client already
-speaks one of these protocols:
+Enable the protocol server that matches the external client:
 
 ```yaml
 apiVersion: stable-harness.dev/v1
 kind: Runtime
 spec:
   protocols:
-    agentProtocols:
+    gateway:
       enabled: true
       host: 127.0.0.1
-      port: 8650
-      protocols: [acp, a2a, agui]
+      port: 56789
+    a2a:
+      enabled: true
 ```
 
-The server exposes A2A over HTTP+JSON/SSE, AG-UI over HTTP+SSE, and ACP over a
-documented HTTP JSON-RPC custom transport. Local ACP clients can also launch:
+Stable Harness exposes A2A over HTTP+JSON/SSE, AG-UI over HTTP+SSE, and ACP
+over a documented HTTP JSON-RPC custom transport. Local ACP clients can also
+launch:
 
 ```bash
 stable-harness acp-stdio -w ./workspace
@@ -237,7 +238,7 @@ stable-harness acp-stdio -w ./workspace
 
 The stdio entrypoint accepts one JSON-RPC message per line and reuses the same
 ACP method handler as the HTTP endpoint. See
-[ACP, A2A, and AG-UI Protocol Facades](../protocols/agent-protocols.md).
+[ACP, A2A, and AG-UI Protocol Facades](../protocols/protocol-facades.md).
 
 ## Backend Adapters
 

package/docs/guides/operator-runbook.md

@@ -68,7 +68,8 @@ stable-harness -w ./workspace --agent orchestra --trace "Use the echo tool with
 Start:
 
 ```bash
-stable-harness start -w ./workspace --host 127.0.0.1 --port 8642
+stable-harness start -w ./workspace --host 127.0.0.1
+stable-harness serve -w ./workspace --host 127.0.0.1
 ```
 
 Stop:
@@ -77,10 +78,57 @@ Stop:
 stable-harness stop -w ./workspace
 ```
 
-Use `STABLE_HARNESS_OPENAI_HOST`, `STABLE_HARNESS_OPENAI_PORT`, and
-`STABLE_HARNESS_OPENAI_API_KEY` when the server is managed by an environment
+The default daemon port is the Stable Harness gateway on `56789`. Use
+`STABLE_HARNESS_HOST`, `STABLE_HARNESS_PORT`, `STABLE_HARNESS_GATEWAY_PORT`,
+and `STABLE_HARNESS_API_KEY` when the server is managed by an environment
 wrapper.
 
+## Inspect System Data
+
+Daemon mode uses the workspace-local `runtime.dataRoot`, defaulting to
+`.stable-harness`, for runtime-owned system data:
+
+```text
+.stable-harness/
+  manifest.json
+  control-plane/stable.sqlite
+  artifacts/
+  memory/langmem/langmem.sqlite
+  checkpoints/deepagents.sqlite
+  store/deepagents.sqlite
+  langgraph/
+  logs/
+  tmp/
+```
+
+`control-plane/stable.sqlite` is the daemon default for runs, events, sessions,
+approvals, queue state, recovery intents, and artifact metadata. Artifact
+content is stored under `artifacts/`. JSON stores remain available for tests,
+debugging, import, and export, but they are not the daemon default because they
+rewrite whole documents and do not provide indexed concurrent control-plane
+queries.
+
+Inspect the resolved paths:
+
+```bash
+stable-harness data inspect -w ./workspace
+```
+
+Run non-destructive diagnostics:
+
+```bash
+stable-harness data doctor -w ./workspace
+```
+
+Clean only generated temporary files:
+
+```bash
+stable-harness data gc -w ./workspace
+```
+
+`langgraph/` is a generated compatibility projection for LangGraph and
+LangSmith clients. It is not the Stable Harness source of truth.
+
 ## Read Runtime Evidence
 
 Prefer structured runtime evidence over final prose:

package/docs/guides/runtime-governance-proof.md

@@ -23,8 +23,8 @@ stable-harness agent render orchestra -w ./examples/minimal-deepagents
 Start the protocol facade:
 
 ```bash
-stable-harness start -w ./examples/minimal-deepagents --port 8642
-curl -fsS http://127.0.0.1:8642/v1/models
+stable-harness start -w ./examples/minimal-deepagents
+curl -fsS http://127.0.0.1:56789/protocols/openai/v1/models
 ```
 
 Package the same workspace as a portable Docker runtime:
@@ -32,8 +32,8 @@ Package the same workspace as a portable Docker runtime:
 ```bash
 stable-harness build --target docker -w ./examples/minimal-deepagents --output ./dist/workspace-docker
 cd ./dist/workspace-docker
-STABLE_HARNESS_PORT=8642 docker compose up --build
-curl -fsS http://127.0.0.1:8642/v1/models
+STABLE_HARNESS_PORT=56789 docker compose up --build
+curl -fsS http://127.0.0.1:56789/protocols/openai/v1/models
 ```
 
 ## What The Proof Shows

package/docs/guides/workspace-authoring.md

@@ -35,9 +35,9 @@ spec:
     defaultAgentId: orchestra
   protocols:
     inProcess: true
-    openaiCompatible:
+    gateway:
       host: 127.0.0.1
-      port: 8642
+      port: 56789
 ```
 
 Keep runtime config typed and explicit. Do not route by prompt keywords, tool

package/docs/guides/workspace-docker-build.md

@@ -40,7 +40,7 @@ http://127.0.0.1:3000/v1
 Use another host port if `3000` is already allocated:
 
 ```bash
-STABLE_HARNESS_PORT=8642 docker compose up --build
+STABLE_HARNESS_PORT=56789 docker compose up --build
 ```
 
 ## Persistence
@@ -58,12 +58,12 @@ application that defines it.
 ## Versioning
 
 The Dockerfile pins the Stable Harness package version used by the build
-command. It installs that package into an isolated `/stable-runtime` npm project
+command. It installs that package into an isolated `/stable-harness-runtime` npm project
 so runtime dependencies resolve like a normal application install:
 
 ```dockerfile
 ARG STABLE_HARNESS_PACKAGE=stable-harness@<current package version>
-WORKDIR /stable-runtime
+WORKDIR /stable-harness-runtime
 RUN npm init -y && npm install "${STABLE_HARNESS_PACKAGE}"
 ```
 

package/docs/memory/0.1.0-memory-design.zh.md

@@ -436,8 +436,17 @@ easynet:user:e2e-user:user
 
 ```text
 config/memory/                         # Memory YAML
+.stable-harness/manifest.json          # system data manifest
+.stable-harness/control-plane/stable.sqlite
+                                        # runs/events/sessions/approvals/queue/artifact metadata
 .stable-harness/memory/langmem/         # generated LangMem data
 .stable-harness/memory/langmem/e2e/     # disposable E2E SQLite
+.stable-harness/checkpoints/deepagents.sqlite
+                                        # DeepAgents checkpoint substrate
+.stable-harness/store/deepagents.sqlite # DeepAgents/LangGraph Store substrate
+.stable-harness/artifacts/              # artifact content files
+.stable-harness/langgraph/              # generated LangGraph compatibility projection
+.stable-harness/tmp/                    # generated temporary files
 ```
 
 EasyNet：
@@ -452,6 +461,17 @@ config/memory/                          # Memory YAML
 
 生成数据不提交到 git。
 
+daemon 默认使用 SQLite 作为 control-plane 存储。JSON store 只保留为测试、
+debug、import/export 形态，不作为长期运行 daemon 的默认持久化后端。
+原因是 JSON 文件追加事件时需要重写整份文档，缺少并发事务和索引查询。
+
+LangMem 的 vector 数据属于召回索引，不是唯一事实源。事实源仍然是
+`memory_records`；`memory_vectors` 可以由事实记录和 embedding model 重建。
+本地默认可以使用 SQLite 中的 embedding JSON + cosine search；当 memory
+规模、tenant 数量、artifact/trace recall 规模上升时，可以接入 Qdrant、
+LanceDB、pgvector、Chroma 或 Pinecone provider，但 Stable Harness runtime
+不绑定某一个外部 vector database。
+
 ## SQLite Schema
 
 LangMem SQLite service 当前维护两张表：

package/docs/memory/0.1.0-step-09-deepagents-native-memory.zh.md

@@ -135,7 +135,7 @@ flowchart TD
 - `deepagents.memory` 透传为 `createDeepAgent({ memory })`。
 - `deepagents.skills` 透传为 `createDeepAgent({ skills })`。
 - `deepagents.backend`、`store`、`checkpointer`、`middleware`、`permissions` 透传。
-- 真实 `deepagents@1.9.1` + 远端 Ollama `granite4.1:3b` 默认路径仍通过。
+- 真实 `deepagents@1.9.1` + 远端 Ollama `qwen3.5:4b` 默认路径仍通过。
 
 验证命令：
 

package/docs/memory/0.1.0-step-09-langmem-shaped-provider.zh.md

@@ -165,5 +165,5 @@ npm run test:botbotgo:full
 - `npm test` 通过：18 个 contract tests，7 个 real integration tests。
 - `test:botbotgo:full` 通过：8/8 matrix cases。
 - full matrix 明确使用 EasyNet package-local `node_modules/.bin/botbotgo`。
-- 真实模型路径为 EasyNet 配置的远端 Ollama `granite4.1:3b`。
+- 真实模型路径为 EasyNet 配置的远端 Ollama `qwen3.5:4b`。
 - 覆盖 owner：orchestra、software、qa、ops、release、research、secretary、k8s。

package/docs/memory/0.1.0-step-09-memory-adapter-projection.zh.md

@@ -84,7 +84,7 @@ flowchart TD
 - DeepAgents native memory config 仍透传给 upstream，不产生 prompt 注入。
 - adapter 仍通过 upstream `createDeepAgent` 路径执行。
 - stable runtime 仍不暴露 DeepAgents `/memories/`。
-- 安装 `deepagents@1.9.1` 后，adapter 默认路径真实 import 上游包并使用远端 Ollama `granite4.1:3b` 完成一次真实调用。
+- 安装 `deepagents@1.9.1` 后，adapter 默认路径真实 import 上游包并使用远端 Ollama `qwen3.5:4b` 完成一次真实调用。
 
 验证命令：
 
@@ -109,8 +109,8 @@ npm run test:botbotgo:full
 - `npm test` 通过：18 个 contract tests，7 个 real integration tests。
 - `test:botbotgo:full` 通过：8/8 matrix cases。
 - full matrix 明确使用 EasyNet package-local `node_modules/.bin/botbotgo`。
-- 真实模型路径为 EasyNet 配置的远端 Ollama `granite4.1:3b`。
-- stable-harness 自身测试真实调用 `deepagents@1.9.1` + `@langchain/ollama` + `https://ollama-rtx-4070.easynet.world/`。
+- 真实模型路径为 EasyNet 配置的远端 Ollama `qwen3.5:4b`。
+- stable-harness 自身测试真实调用 `deepagents@1.9.1` + `@langchain/ollama` + `https://ollama-rtx-4070.easynet.world/v1`。
 - 覆盖 owner：orchestra、software、qa、ops、release、research、secretary、k8s。
 - 覆盖真实工具/数据路径：finance stock report、source analysis、disk investigation、Git/GitHub Actions、Kubernetes readonly investigation、CLI routing。
 

package/docs/memory/0.1.0-step-09-memory-contract.zh.md

@@ -153,7 +153,7 @@ npm run test:botbotgo:full
 
 - EasyNet 使用 `file:../stable-harness` 依赖当前本地 runtime。
 - botbotgo full matrix 明确使用 EasyNet package-local `node_modules/.bin/botbotgo`。
-- 模型路径为远端 Ollama `granite4.1:3b`，配置来自 EasyNet `config/models.yaml`。
+- 模型路径为远端 Ollama `qwen3.5:4b`，配置来自 EasyNet `config/models.yaml`。
 - `npm test` 通过：18 个 contract tests，7 个 real integration tests。
 - `test:botbotgo:full` 通过：8/8 matrix cases。
 - 覆盖 owner：orchestra、software、qa、ops、release、research、secretary、k8s。

package/docs/memory/0.1.0-step-09-memory-governance-approval.zh.md

@@ -130,7 +130,7 @@ npm run test:botbotgo:full
 - `npm test` 通过：18 个 contract tests，7 个 real integration tests。
 - `test:botbotgo:full` 通过：8/8 matrix cases。
 - full matrix 明确使用 EasyNet package-local `node_modules/.bin/botbotgo`。
-- 真实模型路径为 EasyNet 配置的远端 Ollama `granite4.1:3b`。
+- 真实模型路径为 EasyNet 配置的远端 Ollama `qwen3.5:4b`。
 - 覆盖 owner：orchestra、software、qa、ops、release、research、secretary、k8s。
 - 覆盖真实工具/数据路径：finance stock report、source analysis、disk investigation、Git/GitHub Actions、Kubernetes readonly investigation、CLI routing。
 

package/docs/memory/0.1.0-step-09-memory-lifecycle-hooks.zh.md

@@ -136,7 +136,7 @@ npm run test:botbotgo:full
 - `npm test` 通过：18 个 contract tests，7 个 real integration tests。
 - `test:botbotgo:full` 通过：8/8 matrix cases。
 - full matrix 明确使用 EasyNet package-local `node_modules/.bin/botbotgo`。
-- 真实模型路径为 EasyNet 配置的远端 Ollama `granite4.1:3b`。
+- 真实模型路径为 EasyNet 配置的远端 Ollama `qwen3.5:4b`。
 - 覆盖 owner：orchestra、software、qa、ops、release、research、secretary、k8s。
 - 覆盖真实工具/数据路径：finance stock report、source analysis、disk investigation、Git/GitHub Actions、Kubernetes readonly investigation、CLI routing。
 

package/docs/memory/0.1.0-step-09-memory-maintenance-boundary.zh.md

@@ -113,6 +113,6 @@ npm run test:botbotgo:full
 - `npm test` 通过：18 个 contract tests，7 个 real integration tests。
 - `test:botbotgo:full` 通过：8/8 matrix cases。
 - full matrix 明确使用 EasyNet package-local `node_modules/.bin/botbotgo`。
-- 真实模型路径为 EasyNet 配置的远端 Ollama `granite4.1:3b`。
+- 真实模型路径为 EasyNet 配置的远端 Ollama `qwen3.5:4b`。
 - 覆盖 owner：orchestra、software、qa、ops、release、research、secretary、k8s。
 - 覆盖真实工具/数据路径：finance stock report、source analysis、disk investigation、Git/GitHub Actions、Kubernetes readonly investigation、CLI routing。

package/docs/memory/0.1.0-step-09-memory-persistence-boundary.zh.md

@@ -105,7 +105,7 @@ npm run test:botbotgo:full
 - `npm test` 通过：18 个 contract tests，7 个 real integration tests。
 - `test:botbotgo:full` 通过：8/8 matrix cases。
 - full matrix 明确使用 EasyNet package-local `node_modules/.bin/botbotgo`。
-- 真实模型路径为 EasyNet 配置的远端 Ollama `granite4.1:3b`。
+- 真实模型路径为 EasyNet 配置的远端 Ollama `qwen3.5:4b`。
 - 覆盖 owner：orchestra、software、qa、ops、release、research、secretary、k8s。
 - 覆盖真实工具/数据路径：finance stock report、source analysis、disk investigation、Git/GitHub Actions、Kubernetes readonly investigation、CLI routing。