thurview 0.1.5 → 0.3.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "thurview",
-  "version": "0.1.5",
+  "version": "0.3.0",
   "description": "Guided, evidence-anchored reviews of agent-written code. A coding agent authors the review; you read, ask, comment and decide in the browser.",
   "license": "MIT",
   "type": "module",
@@ -32,9 +32,18 @@
     "@toon-format/toon": "2.3.1",
     "axi-sdk-js": "0.1.11",
     "diff": "9.0.0",
+    "graphology": "0.26.0",
+    "graphology-communities-louvain": "2.0.2",
     "markdown-it": "15.0.1",
     "open": "11.0.2",
     "shiki": "4.4.3",
+    "tree-sitter-go": "0.25.0",
+    "tree-sitter-java": "0.23.5",
+    "tree-sitter-javascript": "0.25.0",
+    "tree-sitter-python": "0.25.0",
+    "tree-sitter-rust": "0.24.0",
+    "tree-sitter-typescript": "0.23.2",
+    "web-tree-sitter": "0.27.0",
     "yaml": "2.9.0",
     "zod": "4.5.4"
   },
@@ -42,6 +51,7 @@
     "@types/diff": "8.0.0",
     "@types/node": "26.4.1",
     "esbuild": "0.28.2",
+    "graphology-types": "0.24.8",
     "prettier": "3.9.6",
     "tsx": "4.23.13",
     "typescript": "7.0.2",

package/skills/thurview/SKILL.md

@@ -102,11 +102,36 @@ this step.
 
 ### 3. Study the change
 
-Read the whole diff once (`git diff <base> <head>`). Then read each changed
-symbol's callers and callees at head. Trace the main flow through storage,
-network and configuration boundaries. Compare the stated intent (commit
-messages, PR description, the user's own words) with what the code does. The
-gap is the most valuable finding.
+Read the whole diff once (`git diff <base> <head>`). The diff tells you what
+changed. The code graph tells you what it means, which is what the review is
+for. thurview builds it from the pinned commits with tree-sitter and the
+definitions-and-references query each grammar ships, so ask it rather than
+re-deriving structure from hunks:
+
+```sh
+thurview graph impact --review <id>             # symbols touched, edges added and removed, what reaches them, what tests cover them
+thurview graph callers <symbol> --review <id>   # used, or speculative? (--graph base for the old side)
+thurview graph tests-for <symbol> --review <id>
+thurview graph architecture --review <id>       # file clusters with their hubs, the edges between them, the file-level diff
+```
+
+The graph covers TypeScript, JavaScript, Python, Go, Rust and Java; other
+files are absent from it, not empty. References resolve by name, so treat
+`unresolved` as the size of what it could not place, and `<module>` as code
+outside any definition. If `truncated` is true — `truncated.base`/
+`truncated.head` for impact and architecture, which look at both commits; a
+plain `truncated` for callers and tests-for, which look at one — the repo has
+more supported files than the graph could parse, and that answer is a partial
+view — say so rather than treating an empty result as "nothing there".
+
+Spend the review on what those answer: what the change reaches that the diff
+does not show, which boundaries it crosses, what now depends on what, what it left untested. Then compare the stated
+intent (commit messages, PR description, the user's own words) with what the
+code does. The gap is the most valuable finding.
+
+Do not spend the review on naming, formatting, import order, or missing
+defensive checks. Linters, type checkers and `/code-review` catch those, and a
+reader who wanted them would have run those instead.
 
 Read every range you anchor from the pinned commit, not the working tree:
 `git show <head>:<path>` or `git show <base>:<path>`.
@@ -141,6 +166,10 @@ Source worktree: <worktree>
 Base commit: <base>
 Head commit: <head>
 
+Seed the structure from `thurview graph architecture --review <id>` rather
+than guessing it: communities and their hubs become nodes, its edges the
+edges, and its diff the difference between base and head.
+
 Author <dir>/map.yaml: the head structure under nodes/edges and the base
 structure under base. Do not edit review.md or data.yaml. Do not publish.
 Return when `thurview publish --review <id>` reports no map.yaml errors, or
@@ -185,6 +214,7 @@ threads that need you:
   `threads` lists what to address and `wait.decision` the summary. Go to
   step 9.
 - `accepted`: approved. Report and stop.
+- `closed`: the reader ended the review without approving it. Report and stop.
 - `review-dismissed` or `review-deleted`: stop.
 - error `TIMEOUT` (exit code 1, after `--timeout` seconds, default 3600):
   wait again, or report that the reader has not responded.
@@ -219,4 +249,4 @@ Report completion only when all of these hold:
 - The reader has the URL of a published revision.
 - Every `error` diagnostic is resolved.
 - The map is published, or you said why it is not.
-- The review is waiting on the reader, accepted, dismissed or deleted.
+- The review is waiting on the reader, accepted, closed, dismissed or deleted.

package/skills/thurview/references/lifecycle.md

@@ -25,10 +25,11 @@ warns when the branch moved past them.
 | `awaiting-review`        | Reader reads, asks, comments, decides.                   |
 | `awaiting-agent-updates` | Agent addresses threads, resolves them, republishes.     |
 | `accepted`               | Terminal. Cannot be republished.                         |
-| `rejected`               | Terminal.                                                |
+| `closed`                 | Terminal. Ended without approval. Cannot be republished. |
 
 "Ask now" does not change the status. "Submit review" with "Request changes"
-sets `awaiting-agent-updates`; with "Approve" sets `accepted`.
+sets `awaiting-agent-updates`; with "Approve" sets `accepted`; with "Close"
+sets `closed`.
 
 Dismissal is separate: the reader removes the review from the active list and
 `wait` returns `review-dismissed`. A new publication restores it.
@@ -80,7 +81,7 @@ switch between revisions in the browser.
 
 `thurview wait --review <id> [--timeout <s>]` polls the review and returns
 `wait.reason` in `question`, `awaiting-agent-updates`, `accepted`,
-`rejected`, `review-dismissed`, `review-deleted`, with the threads that need
+`closed`, `review-dismissed`, `review-deleted`, with the threads that need
 you. After the timeout it fails with code `TIMEOUT` (exit 1). A question
 already answered by the agent is not reported again.