simjecture 0.6.2__tar.gz → 0.6.3__tar.gz

{simjecture-0.6.2 → simjecture-0.6.3}/.github/workflows/ci.yml

@@ -69,18 +69,12 @@ jobs:
       - name: Install uv
         run: python -m pip install uv
       - name: Install locked environment
-        run: uv sync --frozen --extra tui --extra dsh --extra workspace --extra public --group browser
+        run: uv sync --frozen --extra tui --extra mcp --extra workspace --extra public --group browser
       - name: Install browser test runtime
         run: uv run playwright install --with-deps chromium
       - uses: actions/setup-node@v4
         with:
           node-version: "22.23.1"
-          cache: npm
-          cache-dependency-path: integrations/dsh/package-lock.json
-      - name: Install pinned DSH test runtime
-        run: npm ci --prefix integrations/dsh
-      - name: Test DSH adapter and real MCP profile
-        run: SIMJECTURE_MCP_EXECUTABLE="$PWD/.venv/bin/simjecture-mcp" npm test --prefix integrations/dsh
       - name: Lint
         run: uv run ruff check .
       - name: Test

{simjecture-0.6.2 → simjecture-0.6.3}/.github/workflows/release.yml

@@ -40,18 +40,12 @@ jobs:
             --ro-bind /lib64 /lib64 \
             /usr/bin/true
       - name: Install locked test environment
-        run: uv sync --frozen --extra tui --extra dsh --extra workspace --extra public --group browser
+        run: uv sync --frozen --extra tui --extra mcp --extra workspace --extra public --group browser
       - name: Install browser test runtime
         run: uv run playwright install --with-deps chromium
       - uses: actions/setup-node@v4
         with:
           node-version: "22.23.1"
-          cache: npm
-          cache-dependency-path: integrations/dsh/package-lock.json
-      - name: Install pinned DSH test runtime
-        run: npm ci --prefix integrations/dsh
-      - name: Test DSH adapter and real MCP profile
-        run: SIMJECTURE_MCP_EXECUTABLE="$PWD/.venv/bin/simjecture-mcp" npm test --prefix integrations/dsh
       - name: Lint release source
         run: uv run ruff check .
       - name: Test release source

{simjecture-0.6.2 → simjecture-0.6.3}/CHANGELOG.md

@@ -4,6 +4,88 @@ This project follows semantic versioning. Dates use ISO 8601.
 
 ## Unreleased
 
+## 0.6.3 — 2026-10-20
+
+### Hosted research
+
+- Let signed-in visitors create private workspaces and projects, organize conversations,
+  and share project descriptions and instructions with the assigned agent. Existing
+  conversations stay in Personal, and the same account can access them across browsers.
+- Add a private hosted knowledge library over completed conversation turns. Search,
+  preview and attach exact passages, choose sources per conversation, and let the agent
+  retrieve earlier work. Enforce account/workspace scope on every search and read;
+  deleting a conversation removes its derived search entries. Guest retention and
+  usage limits are unchanged. Hosted visitors cannot connect server folders.
+
+### Breaking changes
+
+- Minimal is the only supported autonomous study workflow. Remove classical/MVP,
+  DSH, structured and frontier launch/resume routes from the CLI, browser and TUI.
+  Keep archived records readable; resume older workflows in their original version.
+- Remove the bundled DSH runtime, profile, Node dependencies, installer option and
+  `dsh` Python extra. The independent low-level MCP bridge uses the `mcp` extra;
+  workspace and knowledge MCP dependencies remain included in their own extras.
+- Keep native CLI agents and the built-in API agent on the minimal workflow. See
+  [migration guidance](docs/how-to/minimal-migration.md) before upgrading older runs.
+
+### Knowledge and research improvements
+
+- Add a local knowledge library shared across workspace conversations. Connect
+  research folders, search text/code/PDF/Office/notebook sources and completed
+  discussions, preview and attach versioned passages, and choose each conversation's
+  sources. Expose read-only CLI/MCP retrieval to external agents and preserve the
+  passages used during preparation in study records. Index locally without an
+  embedding service; keep knowledge retrieval separate from scientific acceptance.
+- Keep a user's navigation choice when a previous conversation is still loading,
+  and close the mobile sidebar when opening another view.
+
+- Connect to an open SOLIDWORKS session with read-only `windows inspect` and
+  commission part rebuild/export on SOLIDWORKS 2025. Launch batch jobs in their
+  own process and bind its PID-specific automation interface. Correct pywin32
+  registration lookup, method dispatch, export argument types and IGES suffixes;
+  retain the actual output filename casing in export receipts. Skip unreadable
+  COM entries, retry temporary startup binding failures, and verify inspected
+  processes against the selected installation with explicit PID selection when
+  several sessions match.
+
+- Let a worker acknowledge the replan supplied at the start of its turn. A new
+  directive can still interrupt an older turn; an already-delivered directive no
+  longer causes an endless two-second interruption and relaunch cycle.
+
+- Remove automatic five-minute worker interruption and the built-in research
+  adapter's default action ceiling. Run director assessments concurrently with
+  workers, preserve their usage accounting, and expose recorded operating settings
+  in the study CLI and browser. Retain deadlines, resource reservations and review
+  gates. Review retries retain provider reasoning mode, and public API agents also
+  use provider output defaults unless their operator specifies a plan limit.
+
+- Preserve complete hashed review sources/results with ranged read-only access
+  instead of rejecting large files or evidence packets. Add paged file, skill,
+  webpage and remote source reads; spool and rotate console logs rather than
+  failing verbose jobs; stream large local artifact downloads.
+
+- Use provider output-token defaults for built-in API workers and reviewers.
+  Remove the implicit 8K worker and 32K reviewer ceilings, honor an explicit
+  provider-profile `max_tokens`, and record output-budget selection in request
+  metadata. Clarify that this API adapter runs independently of legacy DSH.
+
+- Preserve recent API-worker tool history across autonomous research checkpoints.
+  Reaching the action limit now saves an incomplete checkpoint without spending
+  another request on a tool-free summary. Reviewers retain independent contexts.
+  Give the director bounded recent worker activity and a separate count of turns
+  without recorded changes, so distinct inspection commands cannot hide a stall.
+
+- Configure a separate agent backend for the research director and independent
+  reviewers, including an API worker with a native CLI reviewer or the reverse.
+  Save role-specific connections on launch and resume, expose the choice in the
+  study's advanced settings, and retain the hosted service's managed agents.
+  Preserve the exact hypothesis text when preparing direct CLI resumes.
+
+- Keep the configured review backoff after the director parks a research worker.
+  The director's own wait record no longer triggers another model review as if
+  it were new evidence. New plans, results, steering and operational alerts
+  still interrupt deferral.
+
 ## 0.6.2 — 2026-10-16
 
 - Bring the 0.6.2 candidate series into the stable release: project/workspace

{simjecture-0.6.2 → simjecture-0.6.3}/CITATION.cff

@@ -2,8 +2,8 @@ cff-version: 1.2.0
 message: "If you use this software, please cite the exact software version and Git commit."
 title: "Simjecture"
 type: software
-version: 0.6.2
-date-released: 2026-10-16
+version: 0.6.3
+date-released: 2026-10-20
 abstract: >-
   An evidence-governed runtime for autonomous computational experimentation and
   falsification over hypothesis trees.

{simjecture-0.6.2 → simjecture-0.6.3}/CONTRIBUTING.md

@@ -63,10 +63,9 @@ additional dependencies only when they are needed for your contribution.
   General executor qualification additionally requires a dedicated Linux host,
   reserved job accounts, Landlock and libseccomp; the operator examples under
   `scripts/setup_public_*` must be adapted to that host
-- **DSH integration:** use `uv sync --frozen --extra dsh` and the Node version
-  pinned in [CI](.github/workflows/ci.yml), then run `npm ci --prefix integrations/dsh`
-  and `SIMJECTURE_MCP_EXECUTABLE="$PWD/.venv/bin/simjecture-mcp" npm test --prefix integrations/dsh`.
-  See the [adapter guide](integrations/dsh/README.md#adapter-validation)
+- **MCP interfaces:** install `--extra mcp` for the low-level compatibility bridge
+  or `--extra knowledge` for the knowledge library. Run `tests/test_mcp_server.py`
+  and `tests/test_knowledge.py` for protocol changes. DSH is no longer bundled.
 - **Execution backends and scientific runtimes:** follow the relevant capability
   guide and probe the actual backend before testing. See
   [adding a capability](docs/how-to/add-a-capability.md) and

{simjecture-0.6.2 → simjecture-0.6.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: simjecture
-Version: 0.6.2
+Version: 0.6.3
 Summary: Evidence-governed autonomous experimentation and falsification for computational science
 Project-URL: Homepage, https://github.com/tomzhu0225/simjecture
 Project-URL: Documentation, https://drawingsword.com/simjecture/
@@ -27,10 +27,13 @@ Requires-Dist: pandas<4,>=2.2
 Requires-Dist: pillow<13,>=10
 Requires-Dist: pydantic<3,>=2.9
 Requires-Dist: scipy<2,>=1.12
-Provides-Extra: dsh
-Requires-Dist: mcp<3,>=2; extra == 'dsh'
 Provides-Extra: flash-demo
 Requires-Dist: h5py<4,>=3.10; extra == 'flash-demo'
+Provides-Extra: knowledge
+Requires-Dist: mcp<3,>=2; extra == 'knowledge'
+Requires-Dist: pypdf<7,>=6; extra == 'knowledge'
+Provides-Extra: mcp
+Requires-Dist: mcp<3,>=2; extra == 'mcp'
 Provides-Extra: process
 Requires-Dist: psutil<8,>=6; extra == 'process'
 Provides-Extra: public
@@ -43,7 +46,9 @@ Provides-Extra: tui
 Requires-Dist: textual<9,>=2.0; extra == 'tui'
 Provides-Extra: workspace
 Requires-Dist: h5py<4,>=3.10; extra == 'workspace'
+Requires-Dist: mcp<3,>=2; extra == 'workspace'
 Requires-Dist: psutil<8,>=6; extra == 'workspace'
+Requires-Dist: pypdf<7,>=6; extra == 'workspace'
 Requires-Dist: smolagents[openai]<2,>=1.24; extra == 'workspace'
 Description-Content-Type: text/markdown
 
@@ -82,9 +87,13 @@ validation evidence. Missing evidence sends the work back through the loop.
 
 The **answer** policy can also finish with a reviewed falsification of the original
 claim. New browser studies default to answer; the CLI defaults to repair. Both use
-minimal mode by default: flexible agent strategy, recorded experiments and explicit
+the minimal workflow: flexible agent strategy, recorded experiments and explicit
 scientific acceptance. A deadline can leave the question unresolved.
 
+Version 0.6.3 removes classical/MVP, DSH, structured and frontier launch routes.
+Existing records remain readable. See the [migration guide](docs/how-to/minimal-migration.md)
+for new launches and handling older runs.
+
 New minimal studies allow experimental solver development by default. The agent can
 work in a private source fork while preserving the reference installation. Changed
 builds need independent, source-bound qualification before supporting a claim. Turn
@@ -170,6 +179,11 @@ The installer verifies the release bundle and prepares Python and the browser
 workspace. Reopen it with `~/simjecture/start-workspace`. Research files live under
 `~/simjecture/artifacts/projects`. See [installation options](https://drawingsword.com/simjecture/getting-started/installation.html).
 
+In the local workspace, open **Knowledge → Connect folder** to make research files
+searchable across new conversations. The library also indexes completed workspace
+discussions and study reports, with versioned source citations. Indexing needs no
+embedding service. See [the knowledge library guide](docs/getting-started/research-workspace.md#local-knowledge-library).
+
 1. Choose a compatible API model or an authenticated native agent in the workspace.
 2. Discuss your question and prepare a study brief with **Autonomous research**.
 3. Review its hypothesis, evidence requirements, completion policy and budget.

{simjecture-0.6.2 → simjecture-0.6.3}/README.md

@@ -33,9 +33,13 @@ validation evidence. Missing evidence sends the work back through the loop.
 
 The **answer** policy can also finish with a reviewed falsification of the original
 claim. New browser studies default to answer; the CLI defaults to repair. Both use
-minimal mode by default: flexible agent strategy, recorded experiments and explicit
+the minimal workflow: flexible agent strategy, recorded experiments and explicit
 scientific acceptance. A deadline can leave the question unresolved.
 
+Version 0.6.3 removes classical/MVP, DSH, structured and frontier launch routes.
+Existing records remain readable. See the [migration guide](docs/how-to/minimal-migration.md)
+for new launches and handling older runs.
+
 New minimal studies allow experimental solver development by default. The agent can
 work in a private source fork while preserving the reference installation. Changed
 builds need independent, source-bound qualification before supporting a claim. Turn
@@ -121,6 +125,11 @@ The installer verifies the release bundle and prepares Python and the browser
 workspace. Reopen it with `~/simjecture/start-workspace`. Research files live under
 `~/simjecture/artifacts/projects`. See [installation options](https://drawingsword.com/simjecture/getting-started/installation.html).
 
+In the local workspace, open **Knowledge → Connect folder** to make research files
+searchable across new conversations. The library also indexes completed workspace
+discussions and study reports, with versioned source citations. Indexing needs no
+embedding service. See [the knowledge library guide](docs/getting-started/research-workspace.md#local-knowledge-library).
+
 1. Choose a compatible API model or an authenticated native agent in the workspace.
 2. Discuss your question and prepare a study brief with **Autonomous research**.
 3. Review its hypothesis, evidence requirements, completion policy and budget.

{simjecture-0.6.2 → simjecture-0.6.3}/docs/concepts/architecture.md

@@ -1,7 +1,7 @@
 # How the system fits together
 
 Simjecture separates the agent's freedom to investigate from the authority to
-accept a scientific claim. The current default is **minimal** mode, available
+accept a scientific claim. The supported workflow is **minimal** mode, available
 through the browser workspace and headless launcher.
 
 ## Human direction and study state
@@ -76,9 +76,10 @@ establishes a scientific answer.
 
 ## Other workflows and historical records
 
-Structured and frontier workflows remain explicit launcher choices. DSH and the
-legacy campaign kernel retain their own contracts and role interfaces. Existing
-studies keep their recorded mode; they are not migrated silently.
+Minimal is the only supported study workflow. Classical/MVP, DSH, structured
+and frontier launch routes are removed. Existing records can still be inspected;
+resume those workflows only in their original environment. See
+[migration](../how-to/minimal-migration.md).
 
 The [classic architecture](../archive/classic-architecture.md) and
 [classic evidence model](../archive/classic-evidence-and-claims.md) explain those

{simjecture-0.6.2 → simjecture-0.6.3}/docs/getting-started/installation.md

@@ -103,32 +103,19 @@ Overview, and send your first message. **Connections** is only needed
 to add an optional compatible API endpoint and key. See the
 [research workspace guide](research-workspace.md) for a complete first investigation.
 
-New studies default to minimal mode. Install and log in with Codex GLM, Codex,
+Minimal is the only supported study workflow. Install and log in with Codex GLM, Codex,
 Grok or AGY independently, then choose that backend and model in the launcher.
 Simjecture does not redistribute those CLIs or subscriptions. Native tools remain
 available; numerical execution defaults to Bubblewrap, with explicit cooperative PRoot
 selection available on supported restricted hosts. No DeepSeek
 API key or DSH install is required for native-agent studies.
 
-## Legacy API credentials
+## Built-in API credentials
 
-Credentials are process-local harness inputs. They are never mounted into the
-agent workspace.
-
-```bash
-export DEEPSEEK_API_KEY='your-process-local-key'
-```
-
-The official DeepSeek route is selected automatically when this variable is
-present. `ACS_MODEL_PROVIDER=deepseek` makes that choice explicit. See
-`.env.example` for non-secret configuration names; do not place a real key in
-that tracked example.
-
-Official DeepSeek campaigns retain thinking mode by default. For a
-latency-oriented demonstration, `ACS_DEEPSEEK_THINKING=disabled` selects the
-provider's non-thinking mode without changing campaign turn limits, evidence
-contracts, or audit gates. This is an explicit quality/latency tradeoff rather
-than the scientific default.
+Install the `workspace` extra and configure a compatible API through the workspace's
+**Connections** page, or supply a private provider file to `study --backend builtin`.
+Keep credentials out of research files and version control. Native CLI agents use
+their own existing login. See [the workspace guide](research-workspace.md).
 
 ## Optional simulation runtimes
 
@@ -217,8 +204,8 @@ uv run simjecture web demos/gray_scott_counterexample/record --read-only
 
 The default browser entry point is the research workspace. Use a compatible API or
 an installed native CLI. The experiment monitor remains available at `/monitor` and
-can inspect existing study records. The legacy DSH engine is an explicit alternative;
-it is not required for ordinary workspace setup. See [Web interface](web-interface.md).
+can inspect existing study records. DSH has been removed; see
+[migration](../how-to/minimal-migration.md) and [Web interface](web-interface.md).
 
 ## Optional terminal dashboard (maintenance mode)
 

{simjecture-0.6.2 → simjecture-0.6.3}/docs/getting-started/research-workspace.md

@@ -190,21 +190,122 @@ inputs; the research service's existing separate experiment storage limits still
 Tool installations are shared on the machine; a project's inputs and study outputs
 belong to that project.
 
+## Local knowledge library
+
+Open **Knowledge** in the sidebar to search connected research folders, completed
+workspace conversations and study reports. Use **Connect folder** with an absolute
+path on the machine running Simjecture, such as `/home/you/research`. An optional
+name and exclusion patterns help keep generated files out of the library.
+
+![Search results and a versioned passage preview in the local knowledge library](../_static/knowledge/library-desktop.png)
+
+The same controls work in a [narrow browser layout](../_static/knowledge/library-mobile.png).
+These screenshots use sample research notes.
+
+The first version uses local full-text ranking over titles, paths and passages,
+including Chinese search terms. It needs no embedding model or inference API.
+It reads text and code, notebooks without their outputs, PDF text by page, PPTX
+text and speaker notes by slide, and DOCX paragraphs. Install the `workspace`
+extra for the complete interface, or `knowledge` for standalone extraction and MCP:
+
+```bash
+uv sync --extra workspace
+# Standalone knowledge tools:
+uv sync --extra knowledge
+```
+
+**Search** supports content-type and source filters. Select a result to preview
+its source section and version, move between sections, download the unchanged
+original, or copy a citation. **Discuss this** prepares a new conversation with
+the selected passage attached. **Use in conversation** attaches it to the
+conversation you most recently opened. Neither action sends a message or spends
+provider credit. Edit the draft and send when ready.
+
+The **Knowledge sources** control beside each composer defaults to all enabled
+sources in that workspace. Select a subset or select none. The server applies
+this selection to the built-in agent's `knowledge_search` and `knowledge_read`
+tools and to CLI calls scoped with `--conversation`. Native CLI agents receive
+instructions for the same command-line tools. This controls knowledge retrieval;
+trusted local agents may still have independent operating-system file access.
+
+**Sources** displays counts, last-check times, progress and extraction errors.
+The local web server reconciles enabled sources every minute and after restart.
+**Refresh now** queues an immediate check. Changes are processed incrementally;
+deleted files leave current search and uniquely identifiable renamed files keep
+their identity. An unavailable directory retains its last indexed content with
+an unavailable status. Shared folder registrations and their index are reused
+when the same path and exclusion patterns are connected to another workspace;
+different exclusions use separate indexes, while connections and enabled state
+are workspace-specific. Disconnecting a folder removes that workspace's
+access without deleting original files.
+
+Sources are reference context, not accepted scientific evidence. Each read
+exposes a revision and an exact section citation. Agent reads during interactive
+turns and manually attached passages are preserved under the turn's
+`knowledge-context/`. The most recent 32 distinct preparation passages from
+currently allowed sources are copied to an autonomous study's
+`research/project_context/knowledge/references.json`. Editing a connected file
+does not rewrite these records. Deleting a conversation removes its derived
+conversation and study-search records as well.
+
+The index lives in `artifacts/.knowledge/index.sqlite`; it is local user data.
+Extracted versions are retained for existing citations. Conversation ingestion
+uses completed public result events, not provider reasoning or private transport
+history. Hosted public trials do not expose this local-folder interface.
+
+Current limits: 20,000 eligible files per folder; 2 MiB per text/code file and
+64 MiB per PDF/Office file; 500,000 extracted characters and 400 pages/slides per
+document. Truncation and extraction failures appear under Sources. Hidden
+directories, dependency/build folders, symbolic links and common credential
+filenames are skipped. Raw arrays, binary checkpoints and images are not indexed.
+Scanned PDFs need an existing text layer; diagram understanding, OCR and semantic
+embedding search are not included in this first version.
+
+Use **Agent access** to copy a workspace-specific MCP configuration or CLI
+command. MCP exposes read-only `knowledge_search`, `knowledge_read`, and
+`knowledge_status` tools over stdio; it does not start a campaign or grant source
+registration access. Run it on the same machine as the sources:
+
+```bash
+simjecture knowledge --runs-root artifacts --workspace personal status
+simjecture knowledge --runs-root artifacts add /home/you/research
+simjecture knowledge --runs-root artifacts sync
+simjecture knowledge --runs-root artifacts search "magnetic diffusion"
+simjecture knowledge --runs-root artifacts read DOCUMENT_ID --revision REVISION --chunk 0
+simjecture knowledge --runs-root artifacts --workspace personal mcp
+```
+
+CLI and MCP reads work with the browser closed. Run `sync` to refresh while the
+web server is stopped. Local indexing does not upload sources; when a configured
+hosted agent reads a passage, that passage becomes part of its provider context.
+
 ## Research tools
 
-The catalogue uses compact cards for installed and optional tools. **Details**
-opens a separate panel for build variants, paths, setup diagnostics and logs;
-expanding that information does not stretch neighboring cards.
+The catalogue has a search bar, category filter, and **All tools**, **Installed**,
+and **Available to add** filters. Search matches names, descriptions, categories,
+and application or preset names, including connected tools. Click anywhere on a
+card to open its description and available actions. Keyboard users can focus a
+card's **Explore tool** button and press Enter; Escape closes the details dialog.
+Build variants, paths, setup diagnostics and logs stay in that dialog, so opening
+them does not stretch neighboring cards. Filters and open logs survive background
+refreshes.
+
+![Searchable research tool catalogue with category and availability filters](../_static/guides/tool-catalog-desktop.png)
+
+Hosted workspaces show the collection provided by the server administrator, with
+installation and registration controls hidden. Open a card to choose an available
+starter preset or prepare a custom study when the host permits it. Read-only local
+workspaces still allow browsing and searching, with setup actions disabled.
 
 A **Detected** local build is present but still needs a registered capability
 to participate in recorded studies. An unavailable optional managed runtime is
 shown in its scoped setup diagnostics; it does not imply that detected local
 builds are missing. **Connect existing tool** opens the registration form.
-Install, readiness-check, demonstration and agent-assisted setup actions retain
-their existing behavior.
+Install, readiness-check, demonstration and agent-assisted setup actions are in
+the tool's details dialog and retain their existing behavior.
 
 The catalogue reuses `simjecture install` and `simjecture doctor` implementations.
-**Installed** and **Not installed** are separate sections. Installed instruments
+**On your workbench** and **Discover more** are separate sections. Installed instruments
 also show whether a readiness check has passed, failed, or has not yet been run.
 WarpX CPU, EOS and opacity tools have installer buttons. Existing capability descriptors
 in the checkout's `capabilities/`, `.private/*/capabilities*/`, runtime capability

simjecture-0.6.3/docs/getting-started/terminal-ui.md

@@ -0,0 +1,71 @@
+# Terminal interface (maintenance mode)
+
+The [Web interface](web-interface.md) is the primary interactive client. The optional
+Textual dashboard remains available for SSH and browserless use. It launches
+**minimal** studies with a native CLI agent and can inspect historical research
+records. Classical/MVP, DSH, structured and frontier launches are removed; see
+[migration](../how-to/minimal-migration.md) before upgrading an older running study.
+
+## Status and watch
+
+These commands work with the core package and do not require Textual:
+
+```bash
+simjecture status artifacts/my-study
+simjecture status artifacts/my-study --json
+simjecture watch artifacts/my-study
+simjecture watch artifacts/my-study --jsonl
+```
+
+`status` prints a durable snapshot. `watch` follows progress; Ctrl-C stops the
+viewer without cancelling the study. The view reports recorded experiments,
+review state, model usage and time remaining. Unknown usage is not zero, and a
+missing terminal report does not imply that a process is running.
+
+Minimal studies use `research.json`, experiment receipts and
+`research_report.json`. Archived classical records retain their original manifests,
+ledgers, transcripts and reports. Viewing either format does not rewrite it or
+change its scientific conclusions.
+
+## Launch and inspect
+
+```bash
+uv sync --extra tui
+uv run simjecture tui
+uv run simjecture tui artifacts/my-study
+```
+
+The home screen lists recent studies. Its new-study form asks for the hypothesis,
+optional instructions, agent backend and model, execution backend, output directory,
+wall budget and agent watchdog. Minimal is the only workflow; legacy-only memory
+and workspace fields are no longer shown. Review the contract before launching.
+
+The hypothesis and instructions are written to files, and the child process receives
+file paths as arguments. No shell command is built from hypothesis text. The
+saved `study-launch.json` records the contract, and `controller.log` captures output.
+For the built-in API agent, use the browser workspace or
+[headless launcher](../how-to/headless-studies.md).
+
+The dashboard displays hypotheses, validation records and linked evidence. The
+artifact browser and audit ledger remain available for older records. Model prose
+is never treated as scientific acceptance.
+
+## Pause, resume and cancellation
+
+```bash
+simjecture pause artifacts/my-study
+simjecture resume artifacts/my-study --detach
+```
+
+Pause and cancellation require a verified process identity. Minimal studies record
+control requests in the supervisor directory. Resume preserves the saved launch
+contract and original absolute deadline; pausing does not replenish the budget.
+Use [budget extensions](../how-to/extend-study.md) for an explicit increase.
+
+Older classical, DSH, structured and frontier studies cannot resume in this release.
+The dashboard reports migration guidance. Preserve the original environment if you
+need to continue those runs. Cancelling a verified process remains available.
+
+If Textual is missing, `simjecture tui` prints the installation command. Native
+agent credentials and numerical runtimes are configured separately; see
+[installation](installation.md) and [minimal research](../how-to/research-service.md).

{simjecture-0.6.2 → simjecture-0.6.3}/docs/getting-started/web-interface.md

@@ -24,8 +24,8 @@ plots can still be previewed inside the conversation.
 
 The local web interface is the primary human-facing view of a Simjecture
 campaign. It makes the scientific structure visible without replacing the
-durable record. New native studies default to minimal; mode, backend and model
-are separate launch choices. Minimal reads `research.json`, experiment snapshots,
+durable record. New studies use minimal; choose the agent backend and model
+for the launch. Minimal reads `research.json`, experiment snapshots,
 commitments and reviews. Legacy `mvp_manifest.json`, `transcript.jsonl`,
 `hypothesis_ledger.json`, `mvp_report.json`, and artifact provenance remain the
 sources of truth.
@@ -33,8 +33,8 @@ sources of truth.
 Minimal studies show actual agent activity instead of the fixed-stage role strip.
 The header shows mode/backend/model and remaining budget. Long guidance and review
 rationales fold away, with failed reviews expanded. Usage is pending/unavailable
-when the backend has not supplied counters, never assumed to be zero. DSH/API
-remain explicitly labelled legacy routes; incompatible combinations are disabled.
+when the backend has not supplied counters, never assumed to be zero. Archived
+DSH activity remains visible, but its launch route is removed.
 
 ## Open a recorded campaign
 
@@ -160,29 +160,18 @@ page stores a user-supplied compatible API key in a private mode-0600 server fil
 it does not send that key back to the browser.
 
 ```bash
-export DEEPSEEK_API_KEY='your-process-local-key'
 uv run simjecture web --runs-root artifacts
 ```
 
-Campaigns launched in the browser use the same structured
-`operator_input/launch.json` contract as the TUI and CLI. No shell command is
-constructed from hypothesis text. Legacy DSH campaigns use an
-installed DeepSeek Harness profile. DSH owns provider routing,
-retries, conversation compaction, and the resumable model session; Simjecture
-continues to own the hypothesis graph, evidence rules, sandbox, capabilities,
-jobs, and durable scientific record. The DSH session is stored inside the
-campaign and reopened by **Resume** rather than replaced with a blank model
-conversation.
+The advanced form launches minimal research with a native CLI agent using its
+existing login. For the built-in API agent, configure **Connections** and prepare
+a study from a workspace conversation. Both use the same minimal evidence workflow.
 
-The built-in reasoning loop remains available for compatibility and diagnosis:
-
-```bash
-uv run simjecture web --engine native
-```
-
-Viewing and controlling an already recorded campaign does not require Node.js
-or DSH. Only the explicit legacy DSH route needs the
-[DSH deployment guide](../how-to/deepseek-harness.md).
+Browser, TUI and CLI studies preserve `study-launch.json` and the exact operator
+input files. No shell command is constructed from hypothesis text. Minimal is the
+only launch mode; legacy DSH/API, structured and frontier choices are removed.
+Historical records remain inspectable without Node.js or DSH. See
+[migration](../how-to/minimal-migration.md) before upgrading an older running study.
 
 ## Local security boundary
 

{simjecture-0.6.2 → simjecture-0.6.3}/docs/how-to/cylinder-flow.md

@@ -7,7 +7,7 @@ partially saturated solid boundary. NVIDIA Warp and WarpX are separate projects.
 
 ## Install and run
 
-In **Research tools**, select **Cylinder flow → Install**. When the numerical
+In **Research tools**, open the **Cylinder flow** card and select **Install**. When the numerical
 readiness check passes, use **Run cylinder wake** for the starter simulation or
 prepare a custom study in chat. The hosted workspace provides its installed
 runtime; visitors use it without managing global installations.

simjecture-0.6.3/docs/how-to/deepseek-harness.md

@@ -0,0 +1,10 @@
+# DeepSeek Harness route removed
+
+The bundled DSH route, profile and runtime have been removed. Minimal is the only
+supported study workflow. To use a DeepSeek-compatible API, configure the built-in
+API agent through workspace **Connections**, or use `study --backend builtin`
+with a private provider configuration.
+
+See [migration to minimal](minimal-migration.md) for command replacements and
+handling older records. The historical DSH guide remains in the
+[0.6.2 source](https://github.com/tomzhu0225/simjecture/blob/v0.6.2/docs/how-to/deepseek-harness.md).