@paradigma-inc/flywheel 0.1.19 → 0.1.26

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (56) hide show
  1. package/README.md +8 -0
  2. package/package.json +1 -1
  3. package/skills/flywheel/references/experiment-design-protocol.md +1 -1
  4. package/skills/flywheel/references/flywheel-mcp-tool-map.md +32 -35
  5. package/skills/flywheel/setting-up-flywheel/updating-flywheel-mcp.md +8 -0
  6. package/skills/flywheel-auto/SKILL.md +1 -1
  7. package/skills/flywheel-auto/references/ARTIFACTS.md +0 -1
  8. package/skills/flywheel-auto/references/INTERFACES.md +16 -10
  9. package/skills/flywheel-auto/references/experiment-design-protocol.md +1 -1
  10. package/skills/flywheel-auto/references/flywheel-mcp-tool-map.md +32 -35
  11. package/skills/flywheel-lookahead/SKILL.md +22 -15
  12. package/skills/flywheel-lookahead/agents/openai.yaml +3 -3
  13. package/skills/flywheel-lookahead/evals/evals.json +13 -1
  14. package/skills/flywheel-lookahead/references/ARTIFACTS.md +0 -1
  15. package/skills/flywheel-lookahead/references/INTERFACES.md +16 -10
  16. package/skills/flywheel-lookahead/references/flywheel-mcp-tool-map.md +32 -35
  17. package/skills/flywheel-prove/SKILL.md +163 -0
  18. package/skills/flywheel-prove/agents/interface.yaml +4 -0
  19. package/skills/flywheel-prove/assets/pipeline_template/bin/tproof +3 -0
  20. package/skills/flywheel-prove/assets/pipeline_template/bin/tproof.cmd +2 -0
  21. package/skills/flywheel-prove/assets/pipeline_template/logs/.gitkeep +1 -0
  22. package/skills/flywheel-prove/assets/pipeline_template/pyproject.toml +23 -0
  23. package/skills/flywheel-prove/assets/pipeline_template/scripts/smoke_test.cmd +2 -0
  24. package/skills/flywheel-prove/assets/pipeline_template/scripts/smoke_test.sh +3 -0
  25. package/skills/flywheel-prove/assets/pipeline_template/src/tproof/__init__.py +1 -0
  26. package/skills/flywheel-prove/assets/pipeline_template/src/tproof/cli.py +298 -0
  27. package/skills/flywheel-prove/assets/pipeline_template/src/tproof/constants.py +10 -0
  28. package/skills/flywheel-prove/assets/pipeline_template/src/tproof/layout.py +51 -0
  29. package/skills/flywheel-prove/assets/pipeline_template/src/tproof/leanops.py +116 -0
  30. package/skills/flywheel-prove/assets/pipeline_template/src/tproof/runstore.py +58 -0
  31. package/skills/flywheel-prove/assets/pipeline_template/src/tproof/tasking.py +94 -0
  32. package/skills/flywheel-prove/assets/pipeline_template/workspace/prompts/fill_sorries.txt +3 -0
  33. package/skills/flywheel-prove/references/workflow.md +193 -0
  34. package/skills/flywheel-prove/scripts/scaffold_pipeline.py +111 -0
  35. package/skills/flywheel-reproduce/SKILL.md +28 -23
  36. package/skills/flywheel-reproduce/evals/evals.json +7 -1
  37. package/skills/flywheel-reproduce/references/ARTIFACTS.md +0 -1
  38. package/skills/flywheel-reproduce/references/INTERFACES.md +16 -10
  39. package/skills/flywheel-reproduce/references/experiment-design-protocol.md +1 -1
  40. package/skills/flywheel-reproduce/references/flywheel-mcp-tool-map.md +32 -35
  41. package/skills/flywheel-reproduce/references/source-blog.md +35 -0
  42. package/skills/flywheel-reproduce/references/source-generic.md +30 -0
  43. package/skills/flywheel-reproduce/references/source-notes.md +35 -0
  44. package/skills/flywheel-reproduce/references/source-paper.md +89 -0
  45. package/skills/flywheel-reproduce/references/source-wiki.md +36 -0
  46. package/skills/flywheel-to-graph/SKILL.md +26 -21
  47. package/skills/flywheel-to-graph/evals/evals.json +7 -1
  48. package/skills/flywheel-to-graph/references/ARTIFACTS.md +0 -1
  49. package/skills/flywheel-to-graph/references/INTERFACES.md +16 -10
  50. package/skills/flywheel-to-graph/references/flywheel-mcp-tool-map.md +32 -35
  51. package/skills/flywheel-to-graph/references/source-blog.md +34 -0
  52. package/skills/flywheel-to-graph/references/source-generic.md +29 -0
  53. package/skills/flywheel-to-graph/references/source-notes.md +34 -0
  54. package/skills/flywheel-to-graph/references/source-paper.md +85 -0
  55. package/skills/flywheel-to-graph/references/source-wiki.md +35 -0
  56. package/src/cli.mjs +69 -3
@@ -0,0 +1,35 @@
1
+ # Blog Source Reference
2
+
3
+ Use this reference for blog posts and article-style URLs.
4
+
5
+ ## Source Detection
6
+
7
+ Match this reference for sources from common blog/article domains or article-like HTML pages.
8
+
9
+ ## Acquisition
10
+
11
+ - Fetch URL content.
12
+ - Extract main article body and heading structure.
13
+
14
+ ## Decomposition
15
+
16
+ - Short posts: one node is acceptable.
17
+ - Long posts: parent article node plus section children by top-level headings.
18
+ - Complete source decomposition before branch planning.
19
+
20
+ ## Content and Artifacts
21
+
22
+ - Article narrative in node `content`.
23
+ - Summary in `summary`.
24
+ - Supporting media files as artifacts when available.
25
+
26
+ ## Edges
27
+
28
+ - Use hierarchy edges when section children are created.
29
+ - Avoid lateral interpretive edges by default.
30
+
31
+ ## Failure Behavior
32
+
33
+ - If fetch or extraction fails, emit explicit failure and stop.
34
+
35
+ These rules apply only when `$flywheel-reproduce` is the active skill.
@@ -0,0 +1,30 @@
1
+ # Generic Source Reference
2
+
3
+ Use this fallback when source type cannot be confidently matched to paper, wiki, blog, or notes.
4
+
5
+ ## Fallback Trigger
6
+
7
+ Activate this reference only after deterministic and natural-language routing fail to produce a confident match.
8
+
9
+ ## Acquisition
10
+
11
+ - Ask once for missing source location when not recoverable.
12
+ - Perform best-effort extraction from the provided source.
13
+
14
+ ## Decomposition
15
+
16
+ - If top-level headings are available, create parent and section children.
17
+ - If no structural cues are available, create one node with explicit summary of limitations.
18
+ - Complete source decomposition before any validation branch planning.
19
+
20
+ ## Content and Artifacts
21
+
22
+ - Keep extracted narrative in `content`.
23
+ - Attach supporting files as artifacts.
24
+
25
+ ## Failure Behavior
26
+
27
+ - If extraction fails, emit explicit failure and stop.
28
+ - Do not silently produce an underspecified graph.
29
+
30
+ These rules apply only when `$flywheel-reproduce` is the active skill.
@@ -0,0 +1,35 @@
1
+ # Notes Source Reference
2
+
3
+ Use this reference for local note directories such as Obsidian vaults, Logseq exports, or plain markdown note trees.
4
+
5
+ ## Source Detection
6
+
7
+ Match this reference when source is a local notes corpus with many small markdown files.
8
+
9
+ ## Acquisition
10
+
11
+ - Enumerate note files in requested scope.
12
+ - Preserve folder and filename context.
13
+
14
+ ## Decomposition
15
+
16
+ - Default: one note file -> one node.
17
+ - Large note files: split by top-level headings when useful.
18
+ - Complete decomposition before creating validation branches.
19
+
20
+ ## Content and Artifacts
21
+
22
+ - Note text in `content`.
23
+ - Concise note synopsis in `summary`.
24
+ - Embedded local files become artifacts when relevant.
25
+
26
+ ## Edges
27
+
28
+ - Add edges only for durable relationships (hierarchy, prerequisite, dependency).
29
+ - Keep casual backlinks inside markdown.
30
+
31
+ ## Failure Behavior
32
+
33
+ - If note files cannot be read, emit explicit failure with affected paths.
34
+
35
+ These rules apply only when `$flywheel-reproduce` is the active skill.
@@ -0,0 +1,89 @@
1
+ # Paper Source Reference
2
+
3
+ Use this reference when the source is a paper PDF, arXiv page, DOI URL, or direct PDF URL.
4
+
5
+ ## Source Detection
6
+
7
+ Match this reference when any of these are true:
8
+
9
+ - Local file ends with `.pdf`.
10
+ - URL matches `arxiv.org/abs/`.
11
+ - URL matches `arxiv.org/pdf/`.
12
+ - URL is a DOI resolver link (`doi.org/...`).
13
+ - URL is a direct PDF link.
14
+
15
+ ## Acquisition Rules
16
+
17
+ Acquire source content in this order:
18
+
19
+ 1. Local PDF path: read local file bytes and preserve original filename.
20
+ 2. arXiv HTML URL (`arxiv.org/abs/...` or `arxiv.org/html/...`): fetch page body and parse section structure from HTML.
21
+ 3. arXiv PDF URL (`arxiv.org/pdf/...`): download PDF bytes and treat as PDF input.
22
+ 4. DOI URL: resolve final landing page; if it exposes a PDF, download it, otherwise fetch the HTML body.
23
+ 5. Direct PDF URL: download PDF bytes.
24
+
25
+ ## Parsing Rules
26
+
27
+ For PDF text extraction, use parser fallback chain exactly:
28
+
29
+ `PyMuPDF -> pypdfium2 -> pypdf`
30
+
31
+ - Try one parser at a time in that order.
32
+ - Advance to the next parser only when the previous parser fails to extract usable text.
33
+ - If all parsers fail, emit explicit failure and stop. Do not continue with partial guessed structure.
34
+ - If all parsers fail, do not silently produce a single-summary node with a PDF artifact.
35
+
36
+ License note:
37
+
38
+ - PyMuPDF is AGPL-3.0. Artifex offers a commercial license when AGPL terms are not acceptable for downstream distribution.
39
+
40
+ For arXiv HTML parsing, use LaTeXML selector boundaries:
41
+
42
+ - Top-level sections: `section.ltx_section`
43
+ - Top-level headings: `h2.ltx_title`
44
+ - Subsections: `section.ltx_subsection`
45
+ - Subsection headings: `h3.ltx_title`
46
+
47
+ Do not use naive `section` selectors without class filtering.
48
+
49
+ ## Graph Decomposition Contract
50
+
51
+ Default shape:
52
+
53
+ - Create one parent paper node under the requested root.
54
+ - Create one child node per detected top-level paper section.
55
+ - Typical section set includes Abstract, Introduction, Methods, Results, Discussion, Conclusion, and Related Work when present.
56
+
57
+ ## Content Placement Contract
58
+
59
+ - Parent title: paper title.
60
+ - Parent summary: abstract or concise abstract-derived synopsis.
61
+ - Parent content: one-paragraph orientation plus links/references to section children.
62
+ - Child content: narrative for that section only.
63
+
64
+ ## Artifact Contract
65
+
66
+ - Attach the paper PDF artifact to the parent paper node.
67
+ - Attach extracted figures/tables to the section child where they appear, when extraction is feasible.
68
+
69
+ ## Edge Contract
70
+
71
+ - Use hierarchy edges for paper parent -> section children.
72
+ - Do not add lateral interpretive edges (`cites`, `builds-on`, `derived-from`) by default.
73
+
74
+ ## Insight Controller Contract
75
+
76
+ - Single-paper one-shot import: no separate insight control node is created.
77
+ - Multi-paper corpus import: create one insight control node above paper parent nodes.
78
+
79
+ ## Failure Contract
80
+
81
+ - If acquisition fails, return explicit failure with the failing source variant and stop.
82
+ - If all parsers fail, return explicit failure and stop.
83
+ - Do not silently degrade to one summary node plus artifact-only output.
84
+
85
+ ## Reproduce Ordering Contract
86
+
87
+ Finish paper decomposition before validation branch planning. Do not start step-8 branch execution planning until source decomposition is complete.
88
+
89
+ These rules apply only when `$flywheel-reproduce` is the active skill.
@@ -0,0 +1,36 @@
1
+ # Wiki Source Reference
2
+
3
+ Use this reference for markdown wiki-style corpora, README trees, GitHub wiki exports, and Notion markdown exports.
4
+
5
+ ## Source Detection
6
+
7
+ Match this reference when source material is wiki-like markdown content with multiple linked pages.
8
+
9
+ ## Acquisition
10
+
11
+ - Collect all relevant markdown pages in scope.
12
+ - Preserve stable page identifiers from file paths or page titles.
13
+
14
+ ## Decomposition
15
+
16
+ - One stable page or concept maps to one node by default.
17
+ - Large pages may split by top-level headings when that improves readability.
18
+ - Complete graphification before branch planning and execution.
19
+
20
+ ## Content and Artifacts
21
+
22
+ - Main narrative goes into node `content`.
23
+ - Short synopsis goes into `summary`.
24
+ - Supporting files (images, PDFs, data files, notebooks) are node artifacts.
25
+
26
+ ## Edges
27
+
28
+ - Promote only durable semantic relationships to graph edges.
29
+ - Keep incidental hyperlinks in markdown body.
30
+
31
+ ## Failure Behavior
32
+
33
+ - If pages are unreadable or missing, emit explicit failure and stop for that page.
34
+ - Do not silently skip unreadable pages.
35
+
36
+ These rules apply only when `$flywheel-reproduce` is the active skill.
@@ -44,9 +44,10 @@ Before authoring, recover or establish these inputs:
44
44
  1. Resolve the starting context.
45
45
  - Start from the provided source material or explicit source nodes.
46
46
  - Resolve governing nodes in this order: explicit node ids/slugs, explicitly named current context, focused or recently referenced nodes, then ask the user if nothing stable is recoverable.
47
- 2. Classify the input.
47
+ 2. Classify the input through source-type routing.
48
48
  - If it is already a normalized Flywheel graph export, `flywheel_import_subgraph` may be appropriate.
49
- - If it is a markdown/wiki/blog/README corpus, paper, or research note set, port it explicitly with node and artifact tools.
49
+ - Otherwise, route to one source reference in `references/` using the `Source Type Routing` section below.
50
+ - Complete source routing before step 3 node mapping.
50
51
  3. Map each stable page, claim, or concept to a node.
51
52
  - For exact node-mutation shapes, load [references/flywheel-mcp-tool-map.md](references/flywheel-mcp-tool-map.md).
52
53
  - Create a new node with `flywheel_stage_node_create` when needed.
@@ -107,34 +108,38 @@ Canonical contract shape:
107
108
 
108
109
  The `Source corpus` and `Intended scope` lines are the recovery anchors for later authoring passes.
109
110
 
110
- ## Wiki Porting
111
+ ## Source Type Routing
111
112
 
112
- Use this guide when the source material is a markdown wiki, README corpus, or a Karpathy-style LLM wiki that should become Flywheel nodes.
113
+ Route every non-normalized source through one reference file before authoring.
113
114
 
114
- ### Default Mapping
115
+ Routing order:
115
116
 
116
- - One wiki page or stable concept -> one Flywheel node.
117
- - The main page markdown -> the node body/readme field (`content`).
118
- - A short synopsis of the page -> `summary`.
119
- - Images, PDFs, datasets, notebooks, code snippets, and other supporting files -> node artifacts.
120
- - Durable structural relationships -> graph edges.
121
- - Ordinary wiki hyperlinks -> remain markdown links inside the node body.
117
+ 1. Deterministic routing.
118
+ 2. Natural-language routing.
119
+ 3. Generic fallback.
122
120
 
123
- Do not try to preserve the source wiki as a file-first mirror inside Flywheel. Flywheel is node-first.
121
+ Deterministic routing rules:
124
122
 
125
- ### What The Model Should Do
123
+ - Local `.pdf` files -> `references/source-paper.md`
124
+ - `arxiv.org/abs/*`, `arxiv.org/pdf/*`, DOI URLs, direct PDF URLs -> `references/source-paper.md`
125
+ - Markdown files, wiki exports, wiki-like repositories -> `references/source-wiki.md`
126
+ - Blog/article domains and long-form article URLs -> `references/source-blog.md`
127
+ - Local notes trees (Obsidian, Logseq, plain markdown note directories) -> `references/source-notes.md`
126
128
 
127
- Assume the model is expected to perform the port directly unless the user explicitly asks to build a dedicated importer.
129
+ Natural-language routing rule:
128
130
 
129
- That means:
131
+ - If deterministic routing is inconclusive, use user intent from prompt/context
132
+ (for example, "import this paper", "graph this wiki", "convert this blog
133
+ post", "port these research notes") and pick the matching source reference.
130
134
 
131
- 1. Read the source pages and supporting files.
132
- 2. Create or update nodes explicitly.
133
- 3. Write the page body into node `content`.
134
- 4. Publish supporting files as artifacts.
135
- 5. Add only the graph edges that materially help reasoning or navigation.
135
+ Generic fallback rule:
136
136
 
137
- Do not wait for a magical bulk-import surface if the available Flywheel primitives are already enough.
137
+ - If deterministic and natural-language routing are both inconclusive, load
138
+ `references/source-generic.md`.
139
+
140
+ After choosing a source type, load that reference and follow its acquisition,
141
+ parsing, decomposition, artifact, and failure contracts for the rest of the
142
+ authoring pass.
138
143
 
139
144
  ## `flywheel_import_subgraph`
140
145
 
@@ -14,7 +14,13 @@
14
14
  {
15
15
  "id": "claim-bearing-paper-without-execution",
16
16
  "prompt": "Use $flywheel-to-graph to structure this claim-bearing paper into Flywheel nodes and artifacts, but do not run any validation branches yet.",
17
- "expected_output": "Port the paper into nodes, artifacts, and durable edges without hidden execution or compute acquisition, leaving any later empirical validation to `$flywheel-reproduce`.",
17
+ "expected_output": "Create a parent paper node under the requested root, decompose the paper into section child nodes, keep section narratives in each child `content`, attach the PDF to the parent node, and avoid hidden execution or compute acquisition. Do NOT produce a single `insight` node with the PDF as the only artifact.",
18
+ "files": []
19
+ },
20
+ {
21
+ "id": "paper-pdf-decomposes-into-section-children",
22
+ "prompt": "Use $flywheel-to-graph on the paper PDF in the current folder and attach the result under an existing root node.",
23
+ "expected_output": "Create a parent paper node under the root, add one child node per top-level paper section (for example Abstract, Introduction, Methods, Results, Discussion, Conclusion), place each section narrative in child `content`, and attach the PDF artifact on the parent. Do NOT emit a single `insight` node with only a summary and the PDF artifact.",
18
24
  "files": []
19
25
  }
20
26
  ]
@@ -43,7 +43,6 @@ Canonical artifact upload/finalize contract derived from `mcp_contract.py`.
43
43
 
44
44
  - Stale `expected_revision` on mutating calls returns `409` and requires explicit reconciliation.
45
45
  - Reusing an idempotency key with a different payload hash causes `409_conflict`.
46
- - Empirical `completed` commits must satisfy one of: `artifacts_non_empty`, `no_artifacts_reason_non_empty`.
47
46
  - Prepare item payloads must be structured objects; stringified JSON item payloads are invalid.
48
47
  - Uploading metadata wrappers instead of raw file bytes violates the upload contract.
49
48
  - Finalize requires a valid prepared batch token and staged uploads from that batch.
@@ -9,10 +9,10 @@ Canonical reference for Flywheel MCP public interfaces and contract pointers.
9
9
  - Section order:
10
10
  - `quickstart` (Quickstart): Recommended first calls and section read order for onboarding.
11
11
  - `graph` (Graph): Node model, graph topology guidance, and durable behavior rules.
12
- - `lifecycle` (Lifecycle): Lifecycle operations, commit-time validation requirements, and reproducibility guidance.
12
+ - `stage_commit` (Stage and commit): Stage/commit operations, commit-time validation requirements, and reproducibility guidance.
13
13
  - `sharing` (Sharing): Sharing modes, derived visibility, collaborator roles, and query translation.
14
14
  - `artifacts` (Artifacts): Prepare/upload/finalize requirements and artifact type rules.
15
- - `compute` (Compute): Lease ownership, approval session, and budget source semantics.
15
+ - `compute` (Compute): Lease ownership, token-scoped control, approval session, and budget source semantics.
16
16
  - `compute/troubleshooting_v1` (Compute Troubleshooting v1): Provider-specific acquire/retry hints for launch kwargs and request tuning.
17
17
  - `campaign` (Campaign Contract): Campaign projection and budget contracts plus template section pointers.
18
18
 
@@ -50,10 +50,12 @@ Canonical reference for Flywheel MCP public interfaces and contract pointers.
50
50
  - `flywheel_get_campaign_snapshot` (read; scopes: `read`; core surface; binding: `operation`)
51
51
  - `flywheel_list_audit` (read; scopes: `read`; full-surface only; binding: `operation`)
52
52
 
53
- ### Node lifecycle
53
+ ### Node stage and commit
54
54
 
55
- - `flywheel_stage_node_create` (mutating; scopes: `write`; full-surface only; binding: `operation`)
56
- - `flywheel_stage_node_update` (mutating; scopes: `write`; core surface; binding: `operation`)
55
+ - `flywheel_commit_new_node` (mutating; scopes: `write`; full-surface only; binding: `operation`)
56
+ - `flywheel_acquire_stage_lease` (mutating; scopes: `write`; full-surface only; binding: `operation`)
57
+ - `flywheel_heartbeat_stage_lease` (mutating; scopes: `write`; full-surface only; binding: `operation`)
58
+ - `flywheel_release_stage_lease` (mutating; scopes: `write`; full-surface only; binding: `operation`)
57
59
  - `flywheel_commit_node` (mutating; scopes: `write`; core surface; binding: `operation`)
58
60
  - `flywheel_branch_node` (mutating; scopes: `write`; full-surface only; binding: `operation`)
59
61
  - `flywheel_merge_nodes` (mutating; scopes: `write`; full-surface only; binding: `operation`)
@@ -80,6 +82,7 @@ Canonical reference for Flywheel MCP public interfaces and contract pointers.
80
82
  ### Compute and budgets
81
83
 
82
84
  - `flywheel_compute_list_options` (read; scopes: `compute`; core surface; binding: `operation`)
85
+ - `flywheel_compute_funding` (read; scopes: `compute`; core surface; binding: `operation`)
83
86
  - `flywheel_compute_status` (read; scopes: `compute`; core surface; binding: `operation`)
84
87
  - `flywheel_compute_connection` (read; scopes: `compute`; core surface; binding: `operation`)
85
88
  - `flywheel_approval_session_heartbeat` (read; scopes: `compute`; core surface; binding: `operation`)
@@ -132,10 +135,12 @@ Canonical reference for Flywheel MCP public interfaces and contract pointers.
132
135
  - `GET /mcp/nodes/{node_id}/campaign/snapshot` -> `flywheel_get_campaign_snapshot`
133
136
  - `GET /mcp/nodes/{node_id}/audit` -> `flywheel_list_audit`
134
137
 
135
- ### Node lifecycle
138
+ ### Node stage and commit
136
139
 
137
- - `POST /mcp/nodes/stage/create` -> `flywheel_stage_node_create`
138
- - `PATCH /mcp/nodes/{node_id}/stage/update` -> `flywheel_stage_node_update`
140
+ - `POST /mcp/nodes/commit-new` -> `flywheel_commit_new_node`
141
+ - `POST /mcp/nodes/{node_id}/stage/lease/acquire` -> `flywheel_acquire_stage_lease`
142
+ - `POST /mcp/nodes/{node_id}/stage/lease/heartbeat` -> `flywheel_heartbeat_stage_lease`
143
+ - `POST /mcp/nodes/{node_id}/stage/lease/release` -> `flywheel_release_stage_lease`
139
144
  - `POST /mcp/nodes/{node_id}/commit` -> `flywheel_commit_node`
140
145
  - `POST /mcp/nodes/{node_id}/branch` -> `flywheel_branch_node`
141
146
  - `POST /mcp/nodes/merge` -> `flywheel_merge_nodes`
@@ -161,7 +166,8 @@ Canonical reference for Flywheel MCP public interfaces and contract pointers.
161
166
 
162
167
  ### Compute and budgets
163
168
 
164
- - `GET /mcp/nodes/{node_id}/compute/options` -> `flywheel_compute_list_options`
169
+ - `GET /mcp/compute/catalog` -> `flywheel_compute_list_options`
170
+ - `GET /mcp/compute/funding` -> `flywheel_compute_funding`
165
171
  - `GET /mcp/compute/status` -> `flywheel_compute_status`
166
172
  - `GET /mcp/compute/connection` -> `flywheel_compute_connection`
167
173
  - `POST /mcp/approval-sessions/heartbeat` -> `flywheel_approval_session_heartbeat`
@@ -173,7 +179,7 @@ Canonical reference for Flywheel MCP public interfaces and contract pointers.
173
179
  - `POST /mcp/nodes/{root_node_id}/campaign-budgets` -> `flywheel_create_campaign_budget`
174
180
  - `PATCH /mcp/nodes/{root_node_id}/campaign-budgets/{compute_budget_id}` -> `flywheel_update_campaign_budget`
175
181
  - `DELETE /mcp/nodes/{root_node_id}/campaign-budgets/{compute_budget_id}` -> `flywheel_revoke_campaign_budget`
176
- - `POST /mcp/nodes/{node_id}/compute/acquire` -> `flywheel_compute_acquire`
182
+ - `tool-mediated` -> `flywheel_compute_acquire`
177
183
  - `POST /mcp/compute/release` -> `flywheel_compute_release`
178
184
  - `POST /mcp/compute/release-all` -> `flywheel_compute_release_all`
179
185
 
@@ -12,13 +12,6 @@ Flywheel is a graph-based system for tracking research work, decisions, and evid
12
12
  - Node references include immutable `node_id` and optional immutable `slug_name`; prefer communicating both together for human clarity and disambiguation.
13
13
  - Insight nodes should represent conceptual observations (theoretical insights, intuitions, motivations, decision-relevant framing); empirical nodes should represent experiments with explicit hypotheses and measured outcomes.
14
14
  - Graph topology should encode logical/causal relations between concepts and experiments. Avoid defaulting to shallow root-only branching unless work items are truly independent.
15
- - Node lifecycle semantics are interface-agnostic (`stage_node_create`, `stage_node_update`, `commit_node`); MCP tools are one projection of this shared contract.
16
- - Mutating node writes are optimistic-locking operations: read latest state, pass `expected_revision`, and handle `409 conflict` with explicit reconciliation.
17
- - Mutating operations are idempotent; MCP tool transport auto-manages `Idempotency-Key` on mutating tool calls.
18
- - Commit is finalize-only: commit requests require `expected_revision` and may optionally override `summary`; committed node state must still satisfy strict contract (`summary`/`outcome`, `empirical+completed` requires artifacts or `no_artifacts_reason`, `insight` requires non-empty insights).
19
- - When code is involved, pass `repo_url`/`branch_name`/`head_commit_sha` and align git structure with graph topology where practical (without forcing one-to-one mapping).
20
- - Summaries, hypotheses, and artifacts should be reproduction-grade: enough setup, method, evidence, and interpretation for another reader to reproduce or audit results.
21
- - Empirical workflow is hypothesis-driven: launch execution, inspect outcomes, publish evidence artifacts, and commit only after terminal status.
22
15
  - For empirical work, publish evidence with `flywheel_prepare_artifact_uploads`, upload raw file bytes, then `flywheel_finalize_artifact_uploads` before commit.
23
16
  - Artifact metadata records expose a non-empty `title` suitable for display labels; title normalization must never derive from `storage_url`.
24
17
 
@@ -47,11 +40,13 @@ Flywheel is a graph-based system for tracking research work, decisions, and evid
47
40
  - `flywheel_get_campaign_snapshot` (read; scopes: `read`; HTTP: `GET /mcp/nodes/{node_id}/campaign/snapshot`; core surface): Read the current campaign snapshot for a node's root campaign, including configured views and derived records.
48
41
  - `flywheel_list_audit` (read; scopes: `read`; HTTP: `GET /mcp/nodes/{node_id}/audit`; full-surface only): List node MCP audit events.
49
42
 
50
- ### Node lifecycle
43
+ ### Node stage and commit
51
44
 
52
- - `flywheel_stage_node_create` (mutating; scopes: `write`; HTTP: `POST /mcp/nodes/stage/create`; full-surface only): Stage creation of a new Flywheel node.
53
- - `flywheel_stage_node_update` (mutating; scopes: `write`; HTTP: `PATCH /mcp/nodes/{node_id}/stage/update`; core surface): Stage mutable node fields, including content/readme text, with optimistic locking; use `no_artifacts_reason` when empirical completed nodes intentionally have no artifacts.
54
- - `flywheel_commit_node` (mutating; scopes: `write`; HTTP: `POST /mcp/nodes/{node_id}/commit`; core surface): Commit a node with contract validation.
45
+ - `flywheel_commit_new_node` (mutating; scopes: `write`; HTTP: `POST /mcp/nodes/commit-new`; full-surface only): Commit a locally staged new node into canonical storage and return the persisted node.
46
+ - `flywheel_acquire_stage_lease` (mutating; scopes: `write`; HTTP: `POST /mcp/nodes/{node_id}/stage/lease/acquire`; full-surface only): Acquire a session-scoped stage lease for an existing node before local staged edits.
47
+ - `flywheel_heartbeat_stage_lease` (mutating; scopes: `write`; HTTP: `POST /mcp/nodes/{node_id}/stage/lease/heartbeat`; full-surface only): Refresh the active stage lease for the current editing session.
48
+ - `flywheel_release_stage_lease` (mutating; scopes: `write`; HTTP: `POST /mcp/nodes/{node_id}/stage/lease/release`; full-surface only): Release the active stage lease for the current editing session.
49
+ - `flywheel_commit_node` (mutating; scopes: `write`; HTTP: `POST /mcp/nodes/{node_id}/commit`; core surface): Commit an existing node by publishing the caller's staged payload under an active stage lease.
55
50
  - `flywheel_branch_node` (mutating; scopes: `write`; HTTP: `POST /mcp/nodes/{node_id}/branch`; full-surface only): Create a child branch node.
56
51
  - `flywheel_merge_nodes` (mutating; scopes: `write`; HTTP: `POST /mcp/nodes/merge`; full-surface only): Merge nodes with caller-resolved node payload.
57
52
  - `flywheel_add_parent` (mutating; scopes: `write`; HTTP: `POST /mcp/nodes/{node_id}/parents/add`; full-surface only): Attach an additional parent edge to an existing node (keeps node identity, validates against cycles).
@@ -76,21 +71,22 @@ Flywheel is a graph-based system for tracking research work, decisions, and evid
76
71
 
77
72
  ### Compute and budgets
78
73
 
79
- - `flywheel_compute_list_options` (read; scopes: `compute`; HTTP: `GET /mcp/nodes/{node_id}/compute/options`; core surface): List managed compute options allowed for a node. When using grant-backed compute, pass the same compute_grant_id you will use for acquire.
80
- - `flywheel_compute_status` (read; scopes: `compute`; HTTP: `GET /mcp/compute/status`; core surface): Read managed compute lease status for the current user. Lease rows include ownership flags so hosts can distinguish user-owned leases from sponsor-visible campaign leases. When checking a grant-backed lease, reuse the same compute_grant_id passed to acquire.
81
- - `flywheel_compute_connection` (read; scopes: `compute`; HTTP: `GET /mcp/compute/connection`; core surface): Read SSH connection material for an active managed compute lease once flywheel_compute_status indicates the lease is usable. Only leases owned by the current user are connectable. Pass lease_id or node_id to disambiguate when needed.
74
+ - `flywheel_compute_list_options` (read; scopes: `compute`; HTTP: `GET /mcp/compute/catalog`; core surface): List managed compute catalog options. This endpoint is catalog-only and excludes grant/budget funding fields.
75
+ - `flywheel_compute_funding` (read; scopes: `compute`; HTTP: `GET /mcp/compute/funding`; core surface): Read grant-scoped compute funding context for a specific compute_grant_id.
76
+ - `flywheel_compute_status` (read; scopes: `compute`; HTTP: `GET /mcp/compute/status`; core surface): Read managed compute lease status for the current user and current lease_control_token scope. Lease rows include ownership flags so hosts can distinguish user-owned leases from sponsor-visible campaign leases. When checking a grant-backed lease, reuse the same compute_grant_id passed to acquire.
77
+ - `flywheel_compute_connection` (read; scopes: `compute`; HTTP: `GET /mcp/compute/connection`; core surface): Read SSH connection material for an active managed compute lease once flywheel_compute_status indicates the lease is usable. This tool is token-scoped to lease_control_token and only leases owned by the current user are connectable. Pass lease_id or node_id to disambiguate when needed.
82
78
  - `flywheel_approval_session_heartbeat` (read; scopes: `compute`; HTTP: `POST /mcp/approval-sessions/heartbeat`; core surface): Create or refresh a compute-grant approval session for the current MCP host session.
83
- - `flywheel_list_approval_sessions` (read; scopes: `compute`; HTTP: `GET /mcp/approval-sessions`; core surface): List approval sessions visible to the current user. Optionally include grant approval bindings for each session.
84
- - `flywheel_expire_approval_session` (mutating; scopes: `compute`; HTTP: `POST /mcp/approval-sessions/expire`; core surface): Expire the current compute-grant approval session and release its active leases.
85
- - `flywheel_request_compute_grant_approval` (mutating; scopes: `compute`; HTTP: `tool-mediated`; core surface): Request budget approval and return approval_url + request_id for user confirmation before managed compute acquisition.
79
+ - `flywheel_list_approval_sessions` (read; scopes: `compute`; HTTP: `GET /mcp/approval-sessions`; core surface): List approval sessions visible to the current user.
80
+ - `flywheel_expire_approval_session` (mutating; scopes: `compute`; HTTP: `POST /mcp/approval-sessions/expire`; core surface): Expire the current compute-grant approval session context without releasing active leases.
81
+ - `flywheel_request_compute_grant_approval` (mutating; scopes: `compute`; HTTP: `tool-mediated`; core surface): Request budget approval context before managed compute acquisition; branch on response status.
86
82
  - `flywheel_list_compute_grants` (read; scopes: `compute`; HTTP: `GET /mcp/compute/grants`; core surface): List active/exhausted compute grants available to the current user.
87
83
  - `flywheel_list_campaign_budgets` (read; scopes: `compute`; HTTP: `GET /mcp/nodes/{root_node_id}/campaign-budgets`; full-surface only): List campaign compute budgets for a campaign root. Organizer-only management view.
88
84
  - `flywheel_create_campaign_budget` (mutating; scopes: `compute`; HTTP: `POST /mcp/nodes/{root_node_id}/campaign-budgets`; full-surface only): Create an organizer-funded campaign compute budget shared with participants.
89
85
  - `flywheel_update_campaign_budget` (mutating; scopes: `compute`; HTTP: `PATCH /mcp/nodes/{root_node_id}/campaign-budgets/{compute_budget_id}`; full-surface only): Update hard caps or metadata for an organizer-funded campaign compute budget.
90
86
  - `flywheel_revoke_campaign_budget` (mutating; scopes: `compute`; HTTP: `DELETE /mcp/nodes/{root_node_id}/campaign-budgets/{compute_budget_id}`; full-surface only): Revoke an organizer-funded campaign compute budget.
91
- - `flywheel_compute_acquire` (mutating; scopes: `compute`; HTTP: `POST /mcp/nodes/{node_id}/compute/acquire`; core surface): Acquire managed compute for a node with explicit SKU + region and required compute_grant_id (returns accepted/completed lease state only; poll flywheel_compute_status for readiness, not SSH key material). This tool heartbeats and forwards approval_session_id.
92
- - `flywheel_compute_release` (mutating; scopes: `compute`; HTTP: `POST /mcp/compute/release`; core surface): Asynchronously release one managed compute lease by lease_id.
93
- - `flywheel_compute_release_all` (mutating; scopes: `compute`; HTTP: `POST /mcp/compute/release-all`; core surface): Asynchronously release all active managed compute leases for the current user.
87
+ - `flywheel_compute_acquire` (mutating; scopes: `compute`; HTTP: `tool-mediated`; core surface): Acquire managed compute with explicit SKU + region and required compute_grant_id (returns accepted/completed lease state only; poll flywheel_compute_status for readiness, not SSH key material). This tool forwards approval_session_id and optional context_node_id.
88
+ - `flywheel_compute_release` (mutating; scopes: `compute`; HTTP: `POST /mcp/compute/release`; core surface): Asynchronously release one managed compute lease by lease_id within the current lease_control_token scope.
89
+ - `flywheel_compute_release_all` (mutating; scopes: `compute`; HTTP: `POST /mcp/compute/release-all`; core surface): Asynchronously release active managed compute leases in the current lease_control_token scope; set force=true for explicit account-wide cleanup for the current user.
94
90
 
95
91
  ### Contract, audit, and export
96
92
 
@@ -113,15 +109,15 @@ Flywheel is a graph-based system for tracking research work, decisions, and evid
113
109
  ### Safe Node Update
114
110
 
115
111
  1. `flywheel_get_node`: Read latest node state before mutating fields.
116
- 2. `flywheel_stage_node_update`: Stage changes with fresh expected_revision and resolve 409 conflicts explicitly.
117
- 3. `flywheel_commit_node`: Commit once terminal and contract-complete.
112
+ 2. `flywheel_acquire_stage_lease`: Acquire a session-scoped stage lease before editing an existing node locally.
113
+ 3. `flywheel_commit_node`: Commit with `stage_session_id`, `base_committed_revision`, and full `staged_payload` once terminal and contract-complete.
118
114
 
119
115
  ### Empirical Workflow
120
116
 
121
- 1. `flywheel_stage_node_create`: Create a staged node, then set empirical fields before execution.
122
- 2. `flywheel_stage_node_update`: Set `kind=empirical`, `hypothesis`, and summary fields with fresh `expected_revision`.
123
- 3. `flywheel_request_compute_grant_approval`: If compute is needed, request budget approval first. Response status is `approval_required`.
124
- 4. Branch on `flywheel_request_compute_grant_approval.status`: Branch by response status. `approval_required` is a response state, not a request parameter.. if `approval_required` then `present_approval_url_to_user`: Present `approval_url` to the user; the user opens it and confirms budget approval.; `flywheel_list_approval_sessions`: After approval, list approval sessions with include_approvals=true and use the active `compute_grant_id` for the current approval_session_id.
117
+ 1. `flywheel_commit_new_node`: Commit a local staged new node to canonical storage as the first persistence boundary.
118
+ 2. `flywheel_commit_node`: Commit staged empirical fields with `stage_session_id`, `base_committed_revision`, and a full `staged_payload` once the working state is ready to publish.
119
+ 3. `flywheel_request_compute_grant_approval`: If compute is needed, request budget approval context first. Branch on response status.
120
+ 4. Branch on `flywheel_request_compute_grant_approval.status`: Branch by response status (`already_approved`, `approval_required`, `insufficient_credits`).. if `already_approved` then `reuse_compute_grant_id`: Use returned `compute_grant_id` directly for flywheel_compute_acquire.. if `approval_required` then `present_approval_url_to_user`: Present `approval_url` to the user; the user opens it and confirms budget approval.; `flywheel_list_compute_grants`: After approval, list active grants for the current `approval_session_id` and use the returned `compute_grant_id` for acquire.. if `insufficient_credits` then `request_user_credit_top_up`: No `approval_url` is returned. Ask the user to add credits, then retry flywheel_request_compute_grant_approval.
125
121
  5. `flywheel_compute_acquire`: Acquire lease with `compute_grant_id`; include `approval_session_id` from approval response.
126
122
  6. `flywheel_compute_status`: Poll until the active lease is ready; follow `recommended_next_action`.
127
123
  7. `flywheel_launch_execution`: Launch execution once compute and inputs are ready.
@@ -135,19 +131,20 @@ Flywheel is a graph-based system for tracking research work, decisions, and evid
135
131
 
136
132
  - `flywheel_resolve_node_slug`: resolve human-facing slug references. If response status is `ambiguous`, ask the user to confirm the intended node_id before mutating anything.
137
133
  - `flywheel_get_node`: read the current node state before writes.
138
- - `flywheel_stage_node_update`: update in-progress node fields (title/content/summary, kind/outcome/hypothesis/insights/no_artifacts_reason), always with fresh `expected_revision`.
134
+ - `flywheel_acquire_stage_lease`, `flywheel_heartbeat_stage_lease`, `flywheel_release_stage_lease`: coordinate session-scoped local staged edits for an existing node before commit.
139
135
  - `flywheel_get_campaign_snapshot`: read the current derived campaign state for this node's root campaign instead of inferring standings from freeform text.
140
136
  - `flywheel_get_node_sharing`: after sharing writes, verify with flywheel_get_node_sharing before reporting private/shared/public state.
141
- - `flywheel_compute_status`: check first when work may need managed compute (GPU), to detect any active user lease state.
137
+ - `flywheel_compute_status`: check first when work may need managed compute (GPU), using the active lease_control_token from host context (or pass it explicitly).
142
138
  - `flywheel_list_compute_grants`: list active compute grants (funded by user/root budgets) and select one `compute_grant_id` for acquisition.
143
- - `flywheel_request_compute_grant_approval`: request/confirm budget before acquire and choose a budget source (`user` or `root`); this returns `approval_url` + `request_id` when interactive approval is needed.
144
- - `flywheel_compute_connection`: read SSH connection material for the active user lease once status indicates the lease is usable.
145
- - `flywheel_compute_list_options`: use when a lease is needed and no suitable active lease exists, then select explicit provider-qualified `offer_id` (`provider::offer_id`) and `region`. Consider each option's `availability_mode`: `live_capacity` means provider-reported capacity, `allocation_time` means capacity is confirmed only when `flywheel_compute_acquire` attempts provisioning.
146
- - `flywheel_compute_acquire`: provision compute once requirements are clear. This requires a valid `compute_grant_id` and returns lease/provisioning state only (not SSH key material).
147
- - `flywheel_compute_release`: release compute when no longer needed.
148
- - `flywheel_launch_execution`, `flywheel_list_executions`, `flywheel_terminate_execution`: manage execution lifecycle.
139
+ - `flywheel_request_compute_grant_approval`: request/confirm budget before acquire and choose a budget source (`user` or `root`); branch on status (`already_approved`, `approval_required`, `insufficient_credits`).
140
+ - `flywheel_compute_connection`: read SSH connection material for the active user lease once status indicates the lease is usable, scoped by lease_control_token.
141
+ - `flywheel_compute_list_options`: use when a lease is needed and no suitable active lease exists, then select explicit provider-qualified `offer_id` (`provider::offer_id`) and `region`. This is a catalog-only read and excludes grant/budget money fields.
142
+ - `flywheel_compute_funding`: read grant-scoped funding context (`grant_cents`, `remaining_cents`, backing budget fields) for the selected `compute_grant_id` before acquire.
143
+ - `flywheel_compute_acquire`: provision compute once requirements are clear. This requires a valid `compute_grant_id` and returns lease/provisioning state only (not SSH key material). Capture `compute.lease_control_token` from the response for follow-up lease control tools.
144
+ - `flywheel_compute_release`: release compute when no longer needed, scoped by lease_control_token.
145
+ - `flywheel_launch_execution`, `flywheel_list_executions`, `flywheel_terminate_execution`: manage execution status transitions.
149
146
  - `flywheel_prepare_artifact_uploads`: prepare one or more signed raw-file upload requests for concrete deliverables/evidence produced by the work.
150
147
  - `flywheel_finalize_artifact_uploads`: finalize a staged artifact batch and append all uploaded artifacts in one revision bump.
151
148
  - `flywheel_delete_artifact`: remove an accidental/obsolete node artifact.
152
149
  - `flywheel_list_artifacts`, `flywheel_get_artifact`: inspect node artifact metadata (`title` is the display label) and consume `storage_url` for raw artifact bytes only.
153
- - `flywheel_commit_node`: finalize staged node state once terminal and contract-complete (optional summary override only).
150
+ - `flywheel_commit_new_node`, `flywheel_commit_node`: publish the caller's full staged payload for an existing node once terminal and contract-complete; requires an active stage lease and explicit `base_committed_revision`.
@@ -0,0 +1,34 @@
1
+ # Blog Source Reference
2
+
3
+ Use this reference for blog posts and article-style URLs.
4
+
5
+ ## Source Detection
6
+
7
+ Match this reference for sources from common blog/article domains or article-like HTML pages.
8
+
9
+ ## Acquisition
10
+
11
+ - Fetch URL content.
12
+ - Extract main article body and heading structure.
13
+
14
+ ## Decomposition
15
+
16
+ - Short posts: one node is acceptable.
17
+ - Long posts: parent article node plus section children by top-level headings.
18
+
19
+ ## Content and Artifacts
20
+
21
+ - Article narrative in node `content`.
22
+ - Summary in `summary`.
23
+ - Supporting media files as artifacts when available.
24
+
25
+ ## Edges
26
+
27
+ - Use hierarchy edges when section children are created.
28
+ - Avoid lateral interpretive edges by default.
29
+
30
+ ## Failure Behavior
31
+
32
+ - If fetch or extraction fails, emit explicit failure and stop.
33
+
34
+ These rules apply only when `$flywheel-to-graph` is the active skill.
@@ -0,0 +1,29 @@
1
+ # Generic Source Reference
2
+
3
+ Use this fallback when source type cannot be confidently matched to paper, wiki, blog, or notes.
4
+
5
+ ## Fallback Trigger
6
+
7
+ Activate this reference only after deterministic and natural-language routing fail to produce a confident match.
8
+
9
+ ## Acquisition
10
+
11
+ - Ask once for missing source location when not recoverable.
12
+ - Perform best-effort extraction from the provided source.
13
+
14
+ ## Decomposition
15
+
16
+ - If top-level headings are available, create parent and section children.
17
+ - If no structural cues are available, create one node with explicit summary of limitations.
18
+
19
+ ## Content and Artifacts
20
+
21
+ - Keep extracted narrative in `content`.
22
+ - Attach supporting files as artifacts.
23
+
24
+ ## Failure Behavior
25
+
26
+ - If extraction fails, emit explicit failure and stop.
27
+ - Do not silently produce an underspecified graph.
28
+
29
+ These rules apply only when `$flywheel-to-graph` is the active skill.
@@ -0,0 +1,34 @@
1
+ # Notes Source Reference
2
+
3
+ Use this reference for local note directories such as Obsidian vaults, Logseq exports, or plain markdown note trees.
4
+
5
+ ## Source Detection
6
+
7
+ Match this reference when source is a local notes corpus with many small markdown files.
8
+
9
+ ## Acquisition
10
+
11
+ - Enumerate note files in requested scope.
12
+ - Preserve folder and filename context.
13
+
14
+ ## Decomposition
15
+
16
+ - Default: one note file -> one node.
17
+ - Large note files: split by top-level headings when useful.
18
+
19
+ ## Content and Artifacts
20
+
21
+ - Note text in `content`.
22
+ - Concise note synopsis in `summary`.
23
+ - Embedded local files become artifacts when relevant.
24
+
25
+ ## Edges
26
+
27
+ - Add edges only for durable relationships (hierarchy, prerequisite, dependency).
28
+ - Keep casual backlinks inside markdown.
29
+
30
+ ## Failure Behavior
31
+
32
+ - If note files cannot be read, emit explicit failure with affected paths.
33
+
34
+ These rules apply only when `$flywheel-to-graph` is the active skill.