simjecture 0.3.2__tar.gz → 0.5.0__tar.gz

{simjecture-0.3.2 → simjecture-0.5.0}/.github/workflows/ci.yml

@@ -50,7 +50,16 @@ jobs:
       - name: Install uv
         run: python -m pip install uv
       - name: Install locked environment
-        run: uv sync --frozen --extra tui
+        run: uv sync --frozen --extra tui --extra dsh
+      - 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-0.3.2 → simjecture-0.5.0}/.github/workflows/release.yml

@@ -40,7 +40,16 @@ jobs:
             --ro-bind /lib64 /lib64 \
             /usr/bin/true
       - name: Install locked test environment
-        run: uv sync --frozen --extra tui
+        run: uv sync --frozen --extra tui --extra dsh
+      - 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-0.3.2 → simjecture-0.5.0}/.gitignore

@@ -21,3 +21,5 @@ artifacts/
 !demos/agent_roles_validation/record/workspace/artifacts/*.json
 data/
 .private/
+
+integrations/dsh/node_modules/

{simjecture-0.3.2 → simjecture-0.5.0}/CHANGELOG.md

@@ -4,6 +4,46 @@ This project follows semantic versioning. Dates use ISO 8601.
 
 ## Unreleased
 
+## 0.5.0 — 2026-09-21
+
+- Default new native-agent studies to minimal, with structured and frontier
+  selectable in the CLI, browser and TUI. Mode and backend are separate choices;
+  DSH/API remain explicitly labelled legacy routes. Existing records retain their mode.
+- Preserve active counterexample search, minimal-change repair rationale,
+  branching hypothesis ancestry and fresh prospective validation in the smaller service.
+- Add live terminal activity, job/review counts and remaining budget; project
+  minimal evidence and hypothesis trees into the browser and terminal interface.
+- Preserve deadlines across pause/resume, surface provider failures as resumable
+  states, and verify numerical-worker identity before cancellation.
+- Make the launch package usable with native agent logins without installing DSH;
+  retain an optional `setup.sh --with-dsh` path.
+- Retain measured limitations: simple-task follow-ups completed faster but used
+  more tokens; the previous hard FLASH trial did not establish superiority to a plain agent.
+
+## 0.4.0 — 2026-09-20
+
+- Updated the bundled DSH integration to the tested `0.1.5-rc.2` runtime,
+  including current session, tool-call, prompt and structured-output APIs.
+  The Python release and DSH bundle now share version `0.4.0`.
+- Enabled native web search, page retrieval, shell, filesystem, skills,
+  workflows and configured plugins for research agents. Scientific calls retain
+  role restrictions; independent judges retain their frozen-case isolation.
+- Separated the native research workspace from authoritative campaign files.
+  Upgrades start a fresh `.research-v1` conversation, reconcile existing kernel
+  state, and preserve older conversations. Built-in DSH file/shell policies
+  protect records; arbitrary installed plugins remain trusted process code.
+- Added `read_workspace_image` for bounded PNG, JPEG and WebP image inspection
+  through MCP, with dimensions and content hashes. Inspection is not evidence
+  acceptance and requires a vision-capable model.
+- Extended the Simote scientific tool catalog for image reads. Matching Simote
+  changes add AGY workers and judges; Simote remains a separate optional app.
+- Added real-runtime DSH tests for role isolation, context pruning, V2/V3 log
+  migration, durable-job recovery and complete CLI/MCP pause/resume behavior.
+- Verified native search, fetch, file writing and image inspection with real
+  DeepSeek API calls on non-simulation research tasks.
+- Added a Linux launch bundle containing the wheel, source distribution,
+  matching DSH profile, setup/launch scripts and SHA-256 checksums.
+
 ## 0.3.2 — 2026-09-09
 
 - Added `simjecture-call`, a one-shot JSON transport for trusted supervisors,

{simjecture-0.3.2 → simjecture-0.5.0}/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: 0.3.2
-date-released: 2026-09-09
+version: 0.5.0
+date-released: 2026-09-21
 abstract: >-
   An evidence-governed runtime for autonomous computational experimentation and
   falsification over hypothesis trees.

{simjecture-0.3.2 → simjecture-0.5.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: simjecture
-Version: 0.3.2
+Version: 0.5.0
 Summary: Evidence-governed autonomous experimentation and falsification for computational science
 Project-URL: Homepage, https://github.com/tomzhu0225/simjecture
 Project-URL: Documentation, https://github.com/tomzhu0225/simjecture/tree/main/docs
@@ -24,6 +24,7 @@ Requires-Dist: httpx<1,>=0.27
 Requires-Dist: matplotlib<4,>=3.8
 Requires-Dist: numpy<3,>=1.26
 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
@@ -49,9 +50,9 @@ writes its own experiments and diagnostics, and searches deliberately for the
 simplest result that survives independent verification. The harness controls
 what may count as evidence.
 
-> **Research preview:** version 0.3 extends the infrastructure with a
-> license-safe FLASH resistive-MHD capability and an explicitly bounded
-> island-coalescence audit. It does not claim to solve arbitrary
+> **Research preview:** version 0.5 defaults native-agent studies to minimal mode,
+> adds live terminal progress and shared browser/TUI mode selection, and retains
+> the license-safe FLASH capability and bounded scientific audits. It does not claim to solve arbitrary
 > scientific prose, establish truth about nature from one simulator, or replace
 > independent scientific review.
 
@@ -62,6 +63,23 @@ that have a sharp question and a checkable instrument. The 0.3 preview records
 both the origin domain and that generalization, and adds a fluid-MHD campaign
 whose unresolved repair state is preserved rather than hidden.
 
+## Start an agent-owned study
+
+New native-agent studies default to **minimal**: the agent chooses its approach,
+while Simjecture preserves evidence, counterexamples and hypothesis ancestry.
+
+```bash
+simjecture study --campaign ./runs/my-study \
+  --hypothesis-file hypothesis.txt --instructions-file instructions.md \
+  --backend codex-glm --model glm-5.3 --wall-seconds 3600
+```
+
+Use `--mode structured` or `--mode frontier` to select another workflow. Resuming
+keeps the recorded mode and deadline. See [minimal-mode usage](docs/how-to/research-service.md)
+for the scientific rules, backend choices and explicit legacy DSH/API routes.
+The measured hard-task benchmark does not establish a completion advantage over
+a plain agent; this default reflects the preferred agent-owned design.
+
 ## The core idea: AI low-hanging fruit
 
 Scientific problems are not uniformly difficult. A candidate may be buried in
@@ -289,11 +307,11 @@ the human-first projection.
 
 ## DeepSeek Harness integration
 
-The v0.2.2 integration moves the model-facing research loop to DeepSeek Harness
+The v0.4.0 integration moves the model-facing research loop to DeepSeek Harness
 without moving scientific authority out of Simjecture. DSH owns the provider,
 conversation, retry policy, compaction, and resumable session. A native MCP
-boundary implements 22 typed endpoints. A compact persistent Lead Scientist
-sees seven coordination tools and delegates claim work to fresh, scoped
+boundary implements 23 typed endpoints. A compact persistent Lead Scientist
+sees eight scientific coordination/inspection tools plus native research tools and delegates claim work to fresh, scoped
 Falsifier/Experimenter, Blocker Resolver, and Repair Scientist sessions. A separate fresh,
 tool-free Judge reviews surviving claims. The Python campaign kernel continues to own hypotheses,
 evidence contracts, commissioning, skills, simulation capabilities, sandboxing,

{simjecture-0.3.2 → simjecture-0.5.0}/README.md

@@ -13,9 +13,9 @@ writes its own experiments and diagnostics, and searches deliberately for the
 simplest result that survives independent verification. The harness controls
 what may count as evidence.
 
-> **Research preview:** version 0.3 extends the infrastructure with a
-> license-safe FLASH resistive-MHD capability and an explicitly bounded
-> island-coalescence audit. It does not claim to solve arbitrary
+> **Research preview:** version 0.5 defaults native-agent studies to minimal mode,
+> adds live terminal progress and shared browser/TUI mode selection, and retains
+> the license-safe FLASH capability and bounded scientific audits. It does not claim to solve arbitrary
 > scientific prose, establish truth about nature from one simulator, or replace
 > independent scientific review.
 
@@ -26,6 +26,23 @@ that have a sharp question and a checkable instrument. The 0.3 preview records
 both the origin domain and that generalization, and adds a fluid-MHD campaign
 whose unresolved repair state is preserved rather than hidden.
 
+## Start an agent-owned study
+
+New native-agent studies default to **minimal**: the agent chooses its approach,
+while Simjecture preserves evidence, counterexamples and hypothesis ancestry.
+
+```bash
+simjecture study --campaign ./runs/my-study \
+  --hypothesis-file hypothesis.txt --instructions-file instructions.md \
+  --backend codex-glm --model glm-5.3 --wall-seconds 3600
+```
+
+Use `--mode structured` or `--mode frontier` to select another workflow. Resuming
+keeps the recorded mode and deadline. See [minimal-mode usage](docs/how-to/research-service.md)
+for the scientific rules, backend choices and explicit legacy DSH/API routes.
+The measured hard-task benchmark does not establish a completion advantage over
+a plain agent; this default reflects the preferred agent-owned design.
+
 ## The core idea: AI low-hanging fruit
 
 Scientific problems are not uniformly difficult. A candidate may be buried in
@@ -253,11 +270,11 @@ the human-first projection.
 
 ## DeepSeek Harness integration
 
-The v0.2.2 integration moves the model-facing research loop to DeepSeek Harness
+The v0.4.0 integration moves the model-facing research loop to DeepSeek Harness
 without moving scientific authority out of Simjecture. DSH owns the provider,
 conversation, retry policy, compaction, and resumable session. A native MCP
-boundary implements 22 typed endpoints. A compact persistent Lead Scientist
-sees seven coordination tools and delegates claim work to fresh, scoped
+boundary implements 23 typed endpoints. A compact persistent Lead Scientist
+sees eight scientific coordination/inspection tools plus native research tools and delegates claim work to fresh, scoped
 Falsifier/Experimenter, Blocker Resolver, and Repair Scientist sessions. A separate fresh,
 tool-free Judge reviews surviving claims. The Python campaign kernel continues to own hypotheses,
 evidence contracts, commissioning, skills, simulation capabilities, sandboxing,

{simjecture-0.3.2 → simjecture-0.5.0}/docs/conf.py

@@ -3,7 +3,7 @@ from __future__ import annotations
 project = "Simjecture"
 author = "Bowen Zhu"
 copyright = "2026, Bowen Zhu and contributors"
-release = "0.3.0"
+release = "0.5.0"
 
 extensions = [
     "myst_parser",

simjecture-0.5.0/docs/getting-started/first-run.md

@@ -0,0 +1,68 @@
+# First autonomous run
+
+To explore the interface before supplying an API key, replay the committed
+Gray–Scott record. This is read-only and starts neither a model call nor a
+simulation:
+
+```bash
+uv sync --frozen
+uv run python demos/gray_scott_counterexample/verify_record.py
+uv run simjecture web demos/gray_scott_counterexample/record --read-only
+```
+
+See [Recorded Gray–Scott demo](../demos/gray-scott.md) for the scientific result
+and the boundaries of that record.
+
+To start a new study, install and log in to a supported native agent CLI
+(Codex GLM, Codex, Grok or AGY), then open the browser:
+
+```bash
+uv run simjecture web
+```
+
+In **New hypothesis**, choose mode, backend and model separately. **Minimal** is
+the default. It keeps the agent's native tools and planning freedom while recording
+experiments, counterexample searches, hypothesis repairs and independent reviews.
+Structured and frontier modes remain selectable. DSH and direct API routes are
+explicit legacy choices and require their separate provider configuration.
+
+For a terminal run, write a bounded statement in `hypothesis.txt` and its test
+scope, resource constraints and acceptance criteria in `instructions.md`:
+
+```bash
+uv run simjecture study --campaign artifacts/first-study \
+  --hypothesis-file hypothesis.txt --instructions-file instructions.md \
+  --backend codex-glm --model glm-5.3 --wall-seconds 3600
+```
+
+The terminal shows current activity, elapsed/remaining time and experiment/review
+counts. Use `--quiet` to suppress progress. Select `--mode structured` or
+`--mode frontier` when starting a new directory. Existing mode, backend and wall
+deadline are retained on resume.
+
+Minimal records include `research.json`, immutable experiment snapshots,
+prospective repair commitments, independent review receipts and a final
+`research_report.json`. A clean agent exit is only a checkpoint. A supported root
+or an accepted falsification followed by a supported repair completes the study;
+uncertainty stays unresolved. Numerical convergence and physical validity remain
+scientific obligations. See [minimal research](../how-to/research-service.md).
+
+Inspect or control the same directory:
+
+```bash
+uv run simjecture status artifacts/first-study
+uv run simjecture watch artifacts/first-study
+uv run simjecture web artifacts/first-study
+uv run simjecture pause artifacts/first-study
+uv run simjecture resume artifacts/first-study
+```
+
+Pause stops the native agent at the supervisor boundary. Already recorded jobs
+may finish within their existing limits; the wall deadline continues. Cancel or
+deadline exhaustion terminates active recorded jobs. A provider outage pauses
+with the evidence intact rather than inventing a scientific conclusion.
+
+The optional [Terminal interface](terminal-ui.md) offers the same mode/backend
+selection and a dashboard for SSH and headless machines. The legacy `simjecture
+mvp` and [DSH](../how-to/deepseek-harness.md) entry points remain available; their
+workbench/commissioning contracts are unchanged.

{simjecture-0.3.2 → simjecture-0.5.0}/docs/getting-started/installation.md

@@ -17,7 +17,15 @@ uv run simjecture install core
 uv run simjecture doctor --profile core
 ```
 
-## Model credentials
+## Native agent login
+
+New studies default to minimal mode. 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 still uses the Bubblewrap sandbox. No DeepSeek
+API key or DSH install is required for native-agent studies.
+
+## Legacy API credentials
 
 Credentials are process-local harness inputs. They are never mounted into the
 agent workspace.

{simjecture-0.3.2 → simjecture-0.5.0}/docs/getting-started/web-interface.md

@@ -2,10 +2,18 @@
 
 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. `mvp_manifest.json`, `transcript.jsonl`,
+durable record. New native studies default to minimal; mode, backend and model
+are separate launch choices. 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.
 
+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.
+
 ## Open a recorded campaign
 
 The browser interface is included in the core installation and has no Node.js

{simjecture-0.3.2 → simjecture-0.5.0}/docs/how-to/deepseek-harness.md

@@ -1,16 +1,21 @@
 # Run a Simjecture campaign under DSH
 
 The DSH integration is a native-tool profile plus a small resumable driver for
-a durable Simjecture campaign. Version 0.2.2 uses a persistent Lead Scientist
+a durable Simjecture campaign. Version 0.4.0 uses a persistent Lead Scientist
 and fresh, claim-scoped Falsifier, Repair Scientist, and tool-free Judge
 sessions. Fresh workers see bounded kernel state rather than inherited chat.
-The Python MCP process is the only scientific tool surface: it exposes campaign
-state, claims, evidence contracts, skills, literature metadata, workspace files,
-and bounded jobs. Generic shell, filesystem bypass, and unguarded `finish` tools are
-disabled by the profile wherever the host DSH version supports disabling them.
-The profile also sets DSH's approval policy to `never`: all remaining scientific
-side effects already cross the typed CampaignKernel policy and Bubblewrap
-boundary, so no second routine permission loop is inserted around MCP calls.
+Native web search, page fetch, shell, filesystem, skills, workflows and configured
+plugins remain available to research agents. Only authoritative Simjecture tools
+are filtered by scientific role. The Python MCP process governs campaign state,
+contracts, evidence acceptance, and tracked experiments. Its image endpoint
+returns bounded PNG/JPEG/WebP pixels for vision-capable models.
+
+The profile uses DSH's `workspace-write` file policy around a separate research
+directory and approval `never` (no prompts; escalation requests are rejected).
+Experiment execution retains Simjecture's separate Bubblewrap sandbox.
+
+For the upstream version review and migration requirements, see the
+[DSH upgrade assessment](dsh-upgrade-assessment.md).
 
 ## Provision the two runtimes
 
@@ -59,8 +64,8 @@ resulting local bundle into the isolated harness profile:
 ```bash
 SIMJECTURE_DSH_PROFILE="$(simjecture dsh-profile)"
 npm pack "$SIMJECTURE_DSH_PROFILE" --pack-destination /tmp
-dsh plugin --profile simjecture add @deepseek-ai/dsh-headless@0.1.1-rc.2
-dsh plugin --profile simjecture add /tmp/simjecture-dsh-bundle-0.2.2.tgz
+dsh plugin --profile simjecture add @deepseek-ai/dsh-headless@0.1.5-rc.2
+dsh plugin --profile simjecture add /tmp/simjecture-dsh-bundle-0.4.0.tgz
 ```
 
 For a checkout-only development install, use the directory directly when the
@@ -71,11 +76,13 @@ dsh plugin --profile simjecture add "$PWD/integrations/dsh"
 ```
 
 The bundle pins the tested DSH prerelease and MCP-client prerelease in its
-`package.json`. Keep the lockfile produced by `npm pack`/the DSH plugin manager
-with the deployment artifact; do not silently widen either range.
+`package.json`. The source bundle includes a `package-lock.json` for `npm ci` and runtime tests.
+Keep the separate lockfile produced by the DSH plugin manager with the
+deployment artifact; `npm pack` does not create a lockfile. Do not widen the pins.
 
-DSH `0.1.1-rc.2` requires Node.js `^22.19.0` or `>=24.0.0`. Install the DSH CLI
-according to its upstream instructions and verify that `dsh` is on `PATH`
+DSH `0.1.5-rc.2` requires Node.js `^22.19.0` or `>=24.0.0`. Install the DSH CLI
+at that exact version (`npm install -g @deepseek-ai/dsh@0.1.5-rc.2`)
+and verify that `dsh` is on `PATH`
 before installing this isolated profile.
 
 ## Inspect the resolved configuration
@@ -91,8 +98,8 @@ Confirm that the native MCP client is named `simjecture`, starts
 `simjecture-mcp`, carries the `SIMJECTURE_*` environment, and has
 `failOnStartupError: true`. Also confirm `approval.policy: never`, the
 `simjecture` permission preset (`workspace-write` plus `never`), and that the
-generic tool rows are disabled. A startup or handshake failure should stop the
-scientific profile rather than silently fall back to a bypass tool.
+native research tools remain enabled. A startup or handshake failure must stop
+the profile before its first model request.
 
 Completed large execution and workspace-write exchanges stay fully visible for
 one model request. At a later step the profile may replace the balanced exchange
@@ -138,11 +145,12 @@ simjecture-mcp --workspace "$SIMJECTURE_WORKSPACE" \
 
 That command speaks MCP over stdio and expects an MCP client; it is not a
 human-facing REPL. Use DSH's initialize/list-tools/call-tool trace to verify the
-22 explicit MCP endpoints before submitting work. The persistent lead sees seven
-tools: two durable reads, fresh falsification, blocker resolution, and repair
-delegation, independent adjudication, and guarded finalization. Each worker receives a narrower tool
-surface plus a claim-scoped argument guard. Raw judge prepare/commit operations
-are private to the adjudication composite.
+23 explicit MCP endpoints before submitting work. The lead has eight scientific
+coordination/inspection tools alongside native research tools. Workers receive
+role-specific scientific subsets and claim guards, while retaining native tools.
+Native delegated children inherit their parent's scientific restrictions. The
+independent judge sees only its frozen case and DSH's structured-output handoff;
+raw judge prepare/commit operations remain private to the composite.
 
 `observation_sufficient=true` means that a linked artifact satisfies its
 selected prospective contract; it is not a researcher-issued support verdict.
@@ -191,3 +199,65 @@ a `repairs` child; an honestly complete blocker may close as
 the immutable hand-off artifact. If later review finds an error, append a
 corrective audit record rather than editing the original ledger, transcript, or
 report.
+
+## Upgrade an existing deployment
+
+Stop its campaign supervisor and back up the entire campaign directory, including
+`operator_input/dsh_sessions`, before upgrading the CLI and isolated profile.
+Install DSH `0.1.5-rc.2`, replace the headless bundle with the matching version,
+and install Simjecture's `0.4.0` profile using the commands above. Inspect
+`--dump-config` before resuming. Do not mix the `0.1.6` alpha packages into this
+profile. Provider/model selection stays in your DSH configuration.
+
+DSH publishes a new V3 session generation when opening an older supported log
+for writing; it retains the older generation. A rollback must use the complete
+pre-upgrade campaign backup and the matching older runtime/profile. Do not point
+an older DSH runtime at a campaign that has continued under the newer version.
+
+To reproduce adapter validation from this checkout:
+
+```bash
+uv sync --extra dsh
+npm ci --prefix integrations/dsh
+SIMJECTURE_MCP_EXECUTABLE="$PWD/.venv/bin/simjecture-mcp" npm test --prefix integrations/dsh
+uv run pytest -q tests/test_dsh_bundle.py tests/test_dsh_engine.py
+```
+
+The JavaScript tests use the pinned real DSH agent loop, tools, child runtime,
+and persistence, with a deterministic model adapter. The CLI test also starts
+the real Python MCP server and exercises snapshot, finalization refusal,
+pause/resume, and mandatory-startup failure. It skips when the Python executable
+is unavailable. These tests make no paid model requests and run no scientific
+simulation. The test profile uses filesystem polling to avoid host inotify
+limits; this is a test setting, not a change to deployed profiles.
+
+## Research workspace and record boundary
+
+The runner creates a native research directory outside `SIMJECTURE_WORKSPACE`.
+Its default is the sibling `.simjecture-research/<session-id>` directory;
+`SIMJECTURE_DSH_RESEARCH_ROOT` can select another directory. Overlapping research
+and campaign roots are rejected. Native DSH file and shell tools may write in
+this research directory; they cannot directly overwrite campaign files under
+the configured file policy. Network tools remain enabled. Copy experiment source
+into the scientific workspace using `write_workspace_file` before submitting a
+contracted run. Native notes and calculations do not automatically become evidence.
+
+Bundle 0.4.0 uses the deterministic `<launch-session-id>.research-v1` DSH identity.
+This deliberately avoids resuming pre-0.4.0 sessions whose immutable working
+directory was the campaign itself. On first upgrade, the researcher starts a
+fresh conversation and reconciles the existing kernel snapshot; previous logs
+are retained. Later pause/resume uses the new stable identity normally.
+
+This is a boundary for DSH's enforcing built-in tools, not an OS security boundary
+around arbitrary plugin code. Installed extensions are trusted code with the
+DSH process's authority; upstream 0.1.5 workflow workers likewise are not a
+security sandbox. Deploy the scientific service under a separate account or on
+another host if protection against hostile extensions is required. This profile
+does not install that account/service separation.
+
+`read_workspace_image(path)` accepts campaign-relative PNG, JPEG, and WebP files
+up to 4 MiB and 16 megapixels. MCP clients receive actual image content plus
+path, dimensions, byte count and SHA-256 metadata. A vision-capable model is
+required to interpret the pixels. Image reading is non-mutating and does not
+promote an artifact to scientific evidence. The independent judge still receives
+only its frozen adjudication packet, not unrestricted browsing or image tools.

simjecture-0.5.0/docs/how-to/dsh-upgrade-assessment.md

@@ -0,0 +1,73 @@
+# DSH upgrade assessment — 2026-09-20
+
+Migration implemented: Simjecture profile `0.2.3` pins DSH `0.1.5-rc.2`.
+The scientific CampaignKernel and its evidence contract remain unchanged.
+The adapter changes and validation described below replace the earlier
+recommendation to retain `0.1.1-rc.2` pending compatibility work.
+
+The official npm registry currently tags `0.1.5-rc.2` as `latest` and `next`,
+and `0.1.6-alpha.2` as `alpha`. These are all prereleases. The recommended first
+migration target is the default-channel `0.1.5-rc.2`, keeping the alpha work
+separate. Versions were checked with `npm view @deepseek-ai/dsh dist-tags --json`.
+
+## Compatibility findings addressed
+
+The audit inspected upstream tag `dsh-v0.1.5-rc.2`, commit
+`fb2c4b9e698e30edb738bca4cf0618587db7d203`.
+
+| Boundary | Previous adapter | Upstream target | Migration requirement |
+| --- | --- | --- | --- |
+| Session history | `runner.js` and `context-elider.js` read `session.events` | `Session` exposes `snapshotEvents()` and `eventAt()` instead | Migrate history access and exercise resume, summarization, and pruning against real DSH |
+| Surface replacement | `context-elider.js` writes `{ op: 'replace', start, end }` | `SurfaceOp` requires `startSeq` and `endSeq` | Update replacement events and test balanced tool exchanges and preserved original history |
+| Durable sessions | Existing campaign-local DSH logs | Target includes session-format V3 and V2-to-V3 migration | Test upgrade on copies of old logs, including restart and failed migration; retain originals for rollback |
+| Profile composition | Pinned headless profile, disabled generic tools, seven lead tools and scoped workers | Plugin composition and runtime have evolved | Inspect resolved configuration and actual tool catalogs; test mandatory MCP startup failure and every role restriction |
+
+Primary source files:
+
+- [Session history implementation](https://github.com/deepseek-ai/deepseek-harness/blob/dsh-v0.1.5-rc.2/packages/core/session/src/index.ts)
+- [Surface event types](https://github.com/deepseek-ai/deepseek-harness/blob/dsh-v0.1.5-rc.2/packages/core/session/src/types.ts)
+- [V2-to-V3 migration](https://github.com/deepseek-ai/deepseek-harness/blob/dsh-v0.1.5-rc.2/packages/session/session-format-v2-to-v3/src/migration.ts)
+- [Official releases](https://github.com/deepseek-ai/deepseek-harness/releases)
+
+The `0.1.6` release notes additionally describe headless stdin/session/JSON-event
+support, MCP SDK v2, asynchronous agent initialization, deprecation of synchronous
+history readers, and changes to plugin loading. Those features may eventually
+simplify our custom driver, but do not by themselves replace its campaign budget,
+control-file, activity projection, and reconciliation responsibilities. An alpha
+migration should use its asynchronous history API rather than adopting a newly
+deprecated synchronous interface.
+
+## Why the scientific kernel does not need to change
+
+DSH owns model requests, retries, conversation history, compaction, and child
+sessions. CampaignKernel owns contracts, claim disposition, evidence provenance,
+sandboxing, durable simulation jobs, and the deterministic finish gate. The
+confirmed upstream breaks are in the JavaScript adapter's DSH APIs, not the
+scientific MCP contract. No kernel schema migration is justified by these changes.
+
+The version-change acceptance criteria require a real-runtime adapter smoke test in addition
+to the existing bundle tests: fresh campaign, each scoped role, tool-free judge,
+context pruning, pause/resume on migrated logs, interrupted durable job recovery,
+and finalization only through the kernel gate. Existing source-string/fake tests
+alone cannot qualify a DSH version change. Install and test an isolated profile;
+do not migrate active campaign logs in place during evaluation.
+
+## Completed validation
+
+The adapter now uses `snapshotEvents()` / `eventAt()`, V3 surface replacement
+fields, the renamed `ToolCallId` export, and `personaPrefix`. Removed profile
+rows no longer produce missing-entry warnings. Claim guards attach at
+`agent/created`, before a child receives its first prompt. The Blocker Resolver
+allows DSH's schema-validated `structured_output` handoff. The Judge uses the
+supported schema subset and explicitly rejects inconsistent decision/disposition
+pairs before calling the kernel.
+
+The checked-in Node tests run real DSH services with deterministic model output.
+They cover original-history preservation, V2-to-V3 migration, fresh/resumed lead
+sessions, each worker's first-call mutation guard, independent judge tools and
+usage, and job reconciliation without model polling or resubmission. A complete
+CLI/profile/Python-MCP test verifies the exact lead tool catalog, immutable
+snapshot access, rejection of premature finalization, pause/resume, and no model
+request after mandatory MCP startup failure. No paid-provider or long-running
+scientific campaign was used for qualification. Production profiles and existing
+campaign logs were not modified during this work.