simjecture 0.3.1__tar.gz → 0.3.2__tar.gz

{simjecture-0.3.1 → simjecture-0.3.2}/.github/workflows/ci.yml

@@ -83,6 +83,7 @@ jobs:
         run: |
           uv run python demos/gray_scott_counterexample/verify_record.py
           uv run python demos/collisionless_gem_reconnection/verify_record.py
+          uv run python demos/agent_roles_validation/verify_record.py
 
   warpx-instrument:
     runs-on: ubuntu-latest

{simjecture-0.3.1 → simjecture-0.3.2}/.gitignore

@@ -7,6 +7,7 @@ __pycache__/
 .ruff_cache/
 .hypothesis/
 .coverage
+.coverage.*
 htmlcov/
 docs/_build/
 build/
@@ -16,5 +17,7 @@ dist/
 *.sqlite3-shm
 *.sqlite3-wal
 artifacts/
+!demos/agent_roles_validation/record/workspace/artifacts/
+!demos/agent_roles_validation/record/workspace/artifacts/*.json
 data/
 .private/

{simjecture-0.3.1 → simjecture-0.3.2}/CHANGELOG.md

@@ -4,6 +4,23 @@ This project follows semantic versioning. Dates use ISO 8601.
 
 ## Unreleased
 
+## 0.3.2 — 2026-09-09
+
+- Added `simjecture-call`, a one-shot JSON transport for trusted supervisors,
+  including Simote's SSH bridge, with sandbox readiness probing and durable
+  campaign recovery.
+- Added persistent, claim-scoped scientific role assignments with tool filters,
+  operation budgets, idempotent retries, session tracking, and validated handoffs.
+  Independent adjudication continues to use the existing evidence and claim gates.
+- Documented the optional Simote campaign workspace and Codex/Grok role workflow.
+- Included a hashed audit extract and offline verifier for a live Codex/Grok
+  integration campaign. This validates the transport and scientific role workflow
+  on a finite Euler error-bound claim; it is not a new scientific discovery.
+- Bundled built-in capability descriptors in wheels and discover them alongside
+  packaged skills, so installed deployments can use the same resource catalog.
+- Clarified the distinction between authored Python source and upstream
+  observation data when declaring experiment inputs.
+
 ## 0.3.1 — 2026-08-30
 
 - Added equation-of-state capabilities for atoMEC, Singularity-EOS, and M-ANEOS, with a shared `eos` skill that records each code's model limits.

{simjecture-0.3.1 → simjecture-0.3.2}/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: 0.3.1
-date-released: 2026-08-30
+version: 0.3.2
+date-released: 2026-09-09
 abstract: >-
   An evidence-governed runtime for autonomous computational experimentation and
   falsification over hypothesis trees.

{simjecture-0.3.1 → simjecture-0.3.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: simjecture
-Version: 0.3.1
+Version: 0.3.2
 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
@@ -326,6 +326,12 @@ process state or charging stopped-process downtime.
 
 ## Core safeguards
 
+Codex and Grok CLI agents can also work on a shared campaign through Simote's
+claim-scoped role tasks. See [Simote agent roles](docs/how-to/simote-agent-roles.md)
+for assignment, independent judging, and restart recovery.
+The [completed live Codex/Grok validation](demos/agent_roles_validation/) includes
+a hashed audit extract and an offline numerical verifier.
+
 - Network-isolated writable agent workspace with no provider credentials.
 - Typed, single-action model protocol and bounded tool outputs.
 - Prospective evidence contracts with machine-checkable JSON assertions.

{simjecture-0.3.1 → simjecture-0.3.2}/README.md

@@ -290,6 +290,12 @@ process state or charging stopped-process downtime.
 
 ## Core safeguards
 
+Codex and Grok CLI agents can also work on a shared campaign through Simote's
+claim-scoped role tasks. See [Simote agent roles](docs/how-to/simote-agent-roles.md)
+for assignment, independent judging, and restart recovery.
+The [completed live Codex/Grok validation](demos/agent_roles_validation/) includes
+a hashed audit extract and an offline numerical verifier.
+
 - Network-isolated writable agent workspace with no provider credentials.
 - Typed, single-action model protocol and bounded tool outputs.
 - Prospective evidence contracts with machine-checkable JSON assertions.

simjecture-0.3.2/demos/agent_roles_validation/README.md

@@ -0,0 +1,44 @@
+# Live Codex/Grok role validation
+
+This is an actual subscription-CLI integration run, using Simote's Codex and
+Grok drivers, its role-filtered MCP tools, a real loopback SSH connection, and
+Simjecture's Bubblewrap execution/evidence kernel. No model or scientific
+execution was mocked. It deliberately tests a simple false claim, not a new
+scientific discovery.
+
+- Grok falsified the original 0.001 error bound for ten binary64 Euler steps.
+- A fresh Codex Repair Scientist changed only the bound to 0.02 and registered
+  a prospective contract without running an experiment.
+- A fresh Grok Falsifier generated new evidence. An intentionally dropped
+  contract-registration response was recovered by retrying identical arguments
+  and the same operation ID; the durable result did not change.
+- A fresh, tool-free Codex judge accepted the evidence for the repaired finite
+  claim. Simjecture then finalized the record with all five claims closed.
+
+The supervisor was reopened between phases, and the first Falsifier's session
+was also resumed during integration debugging. Failed commissioning controls
+remain in the record alongside their corrected successors. The worker models
+were Grok 4.6 and GPT-6 Astra, using Grok CLI 1.0.13 and Codex CLI 0.153.4.
+
+The original binary64 reference subtraction reports an error of
+0.01920100107144218. The refined enclosure against mathematical exp(-1) is
+[0.019201001071442167, 0.01920100107144217]. Both reject 0.001 and lie below
+0.02. The verifier independently checks the refined interval at 80-digit
+precision:
+
+```bash
+python3 demos/agent_roles_validation/verify_record.py
+```
+
+The configured GPU containers denied Bubblewrap namespaces. Their attempted
+runs remained non-evidence; successful validation used the local Linux CPU
+through SSH instead. Campaign creation now probes this prerequisite first.
+
+`record/` is an explicit audit extract with agent-authored programs, numerical
+artifacts, provenance, claims, and the final report. It omits provider logs,
+native session cursors, worker authentication material, and runtime resource
+caches, and is not a resumable backup. `sha256.json` binds the extract's files.
+The original finalizer report is preserved byte-for-byte; its elapsed time
+describes finalization, not the full external CLI workflow.
+The opt-in reproduction scripts live in the sibling Simote checkout under
+`scripts/live-scientific-campaign.ts` and `scripts/export-scientific-validation.py`.

simjecture-0.3.2/demos/agent_roles_validation/record/artifact_provenance.json

@@ -0,0 +1,295 @@
+{
+  "artifacts": {
+    "artifacts/commission_binary64_ops.json": {
+      "action": "run_python",
+      "action_sha256": "12f3a6d2e52563bc2e2f30017354f912558c5df2a9ce5366f4cd8c4ada7f3b5b",
+      "active_claim_id": "claim_instrument_binary64_euler",
+      "argv_input_coverage_eligible": true,
+      "argv_input_coverage_issues": [],
+      "bytes": 732,
+      "capability": null,
+      "command_argv": [
+        "scripts/commission_binary64_ops.py"
+      ],
+      "evidence_candidate": true,
+      "evidence_eligible": true,
+      "execution_returncode": 0,
+      "execution_stage": null,
+      "execution_succeeded": true,
+      "execution_timed_out": false,
+      "execution_workspace_exceeded": false,
+      "generated_iteration": 11,
+      "input_artifacts": [],
+      "input_artifacts_declared": true,
+      "input_lineage_eligible": true,
+      "input_lineage_issues": [],
+      "job_id": "job_91c7ce244416797b436e49c478ca1e21",
+      "job_status": "succeeded",
+      "mtime_ns": 1788827527587926014,
+      "operation_id": "falsify-1:run-commission-selfcontained",
+      "program_path": "scripts/commission_binary64_ops.py",
+      "program_sha256": "5d0c9bafbfc77eff322f173d2891fea9ce21f7cc0c51956550ca4bebe04163ca"
+    },
+    "artifacts/commission_series_round_002.json": {
+      "action": "run_python",
+      "action_sha256": "d97be630374e7c2c7661f41241e2a66c0fb3a34f74f1a50cfa0379dbac00ae25",
+      "active_claim_id": "claim_instrument_exp_series_round_002",
+      "argv_input_coverage_eligible": true,
+      "argv_input_coverage_issues": [],
+      "bytes": 1404,
+      "capability": null,
+      "command_argv": [
+        "scripts/commission_series_round_002.py"
+      ],
+      "evidence_candidate": true,
+      "evidence_eligible": true,
+      "execution_returncode": 0,
+      "execution_stage": null,
+      "execution_succeeded": true,
+      "execution_timed_out": false,
+      "execution_workspace_exceeded": false,
+      "generated_iteration": 24,
+      "input_artifacts": [],
+      "input_artifacts_declared": true,
+      "input_lineage_eligible": true,
+      "input_lineage_issues": [],
+      "job_id": "job_962a1f549ce362bb9954c0220a2d7df5",
+      "job_status": "succeeded",
+      "mtime_ns": 1788828352023105674,
+      "operation_id": "confirm-1:run-commission-series-round",
+      "program_path": "scripts/commission_series_round_002.py",
+      "program_sha256": "8b8c9af02d5d9a6182f0514ff5eddc1a91af96a8b131b61e6799cd371381a9a6"
+    },
+    "artifacts/commission_series_round_003.json": {
+      "action": "run_python",
+      "action_sha256": "7a9984b2680cb27ecfb51a2eba1020081ab14707b08e79a68b9f39fc272259b3",
+      "active_claim_id": "claim_instrument_exp_series_round_003",
+      "argv_input_coverage_eligible": true,
+      "argv_input_coverage_issues": [],
+      "bytes": 1526,
+      "capability": null,
+      "command_argv": [
+        "scripts/commission_series_round_003.py"
+      ],
+      "evidence_candidate": true,
+      "evidence_eligible": true,
+      "execution_returncode": 0,
+      "execution_stage": null,
+      "execution_succeeded": true,
+      "execution_timed_out": false,
+      "execution_workspace_exceeded": false,
+      "generated_iteration": 32,
+      "input_artifacts": [],
+      "input_artifacts_declared": true,
+      "input_lineage_eligible": true,
+      "input_lineage_issues": [],
+      "job_id": "job_7f79819c8e13ba3cd1352938cfed77d3",
+      "job_status": "succeeded",
+      "mtime_ns": 1788828573825176942,
+      "operation_id": "confirm-1:run-commission-series-round-003",
+      "program_path": "scripts/commission_series_round_003.py",
+      "program_sha256": "0b72f46fb4fe67cea09584cb627127ccafa835558d133359ec3a4562083127a8"
+    },
+    "artifacts/euler_bound_002_prospective.json": {
+      "action": "run_python",
+      "action_sha256": "62f8a11febadba6ed0e7454bd8c53a5740aacee8f91fa02677444d0f1ad7781a",
+      "active_claim_id": "claim_repair_euler_bound_002",
+      "argv_input_coverage_eligible": true,
+      "argv_input_coverage_issues": [],
+      "bytes": 1813,
+      "capability": null,
+      "command_argv": [
+        "scripts/euler_bound_002_prospective.py"
+      ],
+      "evidence_candidate": true,
+      "evidence_eligible": true,
+      "execution_returncode": 0,
+      "execution_stage": null,
+      "execution_succeeded": true,
+      "execution_timed_out": false,
+      "execution_workspace_exceeded": false,
+      "generated_iteration": 35,
+      "input_artifacts": [],
+      "input_artifacts_declared": true,
+      "input_lineage_eligible": true,
+      "input_lineage_issues": [],
+      "job_id": "job_f6e6b3124589bd84968dc6393cb4dd8b",
+      "job_status": "succeeded",
+      "mtime_ns": 1788828618468492097,
+      "operation_id": "confirm-1:run-euler-bound-002-prospective",
+      "program_path": "scripts/euler_bound_002_prospective.py",
+      "program_sha256": "db23372bc313ea46d00362fc004883009d18c098f083b72ef8cce34cb34824e1"
+    },
+    "artifacts/euler_ten_step_error.json": {
+      "action": "run_python",
+      "action_sha256": "626830457356fadce8c511d2eb43a0b782a10a1485dc6d795610e1bfe0f60e13",
+      "active_claim_id": "claim_root",
+      "argv_input_coverage_eligible": true,
+      "argv_input_coverage_issues": [],
+      "bytes": 586,
+      "capability": null,
+      "command_argv": [
+        "scripts/euler_ten_step_error.py"
+      ],
+      "evidence_candidate": true,
+      "evidence_eligible": true,
+      "execution_returncode": 0,
+      "execution_stage": null,
+      "execution_succeeded": true,
+      "execution_timed_out": false,
+      "execution_workspace_exceeded": false,
+      "generated_iteration": 15,
+      "input_artifacts": [],
+      "input_artifacts_declared": true,
+      "input_lineage_eligible": true,
+      "input_lineage_issues": [],
+      "job_id": "job_505e300d5731b62fb2baadbd4d0a50cd",
+      "job_status": "succeeded",
+      "mtime_ns": 1788827571509552217,
+      "operation_id": "falsify-1:run-euler-ten-step",
+      "program_path": "scripts/euler_ten_step_error.py",
+      "program_sha256": "bbb727b065c2c71fdf79242eb2acae124ad0277b968f8685b676c391b98461e4"
+    },
+    "scripts/commission_binary64_ops.py": {
+      "action": "write_file",
+      "action_sha256": "a4cb5e51cbedd9e6c9019b884d05ae15820bbaf425dd54f62dad8de867f90bf2",
+      "active_claim_id": null,
+      "argv_input_coverage_eligible": true,
+      "argv_input_coverage_issues": [],
+      "bytes": 2892,
+      "capability": null,
+      "command_argv": [],
+      "evidence_candidate": false,
+      "evidence_eligible": false,
+      "execution_returncode": null,
+      "execution_stage": null,
+      "execution_succeeded": null,
+      "execution_timed_out": null,
+      "execution_workspace_exceeded": null,
+      "generated_iteration": 3,
+      "input_artifacts": [],
+      "input_artifacts_declared": false,
+      "input_lineage_eligible": true,
+      "input_lineage_issues": [],
+      "job_id": null,
+      "job_status": "succeeded",
+      "mtime_ns": 1788827284965955688,
+      "operation_id": "falsify-1:write-commission-script",
+      "program_path": null,
+      "program_sha256": null
+    },
+    "scripts/commission_series_round_002.py": {
+      "action": "write_file",
+      "action_sha256": "72db565feb5d5eca8de8fda4245d13ff7cc98d7a32f11c39ed2300ddb32f505c",
+      "active_claim_id": null,
+      "argv_input_coverage_eligible": true,
+      "argv_input_coverage_issues": [],
+      "bytes": 5825,
+      "capability": null,
+      "command_argv": [],
+      "evidence_candidate": false,
+      "evidence_eligible": false,
+      "execution_returncode": null,
+      "execution_stage": null,
+      "execution_succeeded": null,
+      "execution_timed_out": null,
+      "execution_workspace_exceeded": null,
+      "generated_iteration": 22,
+      "input_artifacts": [],
+      "input_artifacts_declared": false,
+      "input_lineage_eligible": true,
+      "input_lineage_issues": [],
+      "job_id": null,
+      "job_status": "succeeded",
+      "mtime_ns": 1788828313825265392,
+      "operation_id": "confirm-1:write-commission-script",
+      "program_path": null,
+      "program_sha256": null
+    },
+    "scripts/commission_series_round_003.py": {
+      "action": "write_file",
+      "action_sha256": "6b3457cb2fa10d1316575f7716c5d480b4b0e87dcfdebafd598639bd56e64df9",
+      "active_claim_id": null,
+      "argv_input_coverage_eligible": true,
+      "argv_input_coverage_issues": [],
+      "bytes": 6815,
+      "capability": null,
+      "command_argv": [],
+      "evidence_candidate": false,
+      "evidence_eligible": false,
+      "execution_returncode": null,
+      "execution_stage": null,
+      "execution_succeeded": null,
+      "execution_timed_out": null,
+      "execution_workspace_exceeded": null,
+      "generated_iteration": 30,
+      "input_artifacts": [],
+      "input_artifacts_declared": false,
+      "input_lineage_eligible": true,
+      "input_lineage_issues": [],
+      "job_id": null,
+      "job_status": "succeeded",
+      "mtime_ns": 1788828542844750204,
+      "operation_id": "confirm-1:write-commission-script-003",
+      "program_path": null,
+      "program_sha256": null
+    },
+    "scripts/euler_bound_002_prospective.py": {
+      "action": "write_file",
+      "action_sha256": "46ba99148167766198cd9afe30e6d36b1ef4b3b7716a97c26b30dc6c7e209842",
+      "active_claim_id": null,
+      "argv_input_coverage_eligible": true,
+      "argv_input_coverage_issues": [],
+      "bytes": 4971,
+      "capability": null,
+      "command_argv": [],
+      "evidence_candidate": false,
+      "evidence_eligible": false,
+      "execution_returncode": null,
+      "execution_stage": null,
+      "execution_succeeded": null,
+      "execution_timed_out": null,
+      "execution_workspace_exceeded": null,
+      "generated_iteration": 25,
+      "input_artifacts": [],
+      "input_artifacts_declared": false,
+      "input_lineage_eligible": true,
+      "input_lineage_issues": [],
+      "job_id": null,
+      "job_status": "succeeded",
+      "mtime_ns": 1788828383368273892,
+      "operation_id": "confirm-1:write-euler-prospective-script",
+      "program_path": null,
+      "program_sha256": null
+    },
+    "scripts/euler_ten_step_error.py": {
+      "action": "write_file",
+      "action_sha256": "a29436a6a00d93f270201ad4c324d3a29e597e47f8a263c43493faa9e4cfc02b",
+      "active_claim_id": null,
+      "argv_input_coverage_eligible": true,
+      "argv_input_coverage_issues": [],
+      "bytes": 1406,
+      "capability": null,
+      "command_argv": [],
+      "evidence_candidate": false,
+      "evidence_eligible": false,
+      "execution_returncode": null,
+      "execution_stage": null,
+      "execution_succeeded": null,
+      "execution_timed_out": null,
+      "execution_workspace_exceeded": null,
+      "generated_iteration": 10,
+      "input_artifacts": [],
+      "input_artifacts_declared": false,
+      "input_lineage_eligible": true,
+      "input_lineage_issues": [],
+      "job_id": null,
+      "job_status": "succeeded",
+      "mtime_ns": 1788827509929971511,
+      "operation_id": "falsify-1:write-euler-script",
+      "program_path": null,
+      "program_sha256": null
+    }
+  },
+  "schema_version": "0.1.0"
+}