feat(research): wire tools + research workflow graph (research-workflow §2/§3)

Makes the research feature runnable end-to-end, off by default.

- config: [tools.research] (enabled, searxng_url, max_results, max_fetch_bytes).
- registration: web_search/web_fetch are built into BOTH the default and per-workspace
  tool registries when research.enabled, sharing one HTTP client threaded from Main
  (none built on the static path). Egress stays harness-enforced: web_fetch is T2
  (operator-approved) and the existing NetworkHostRule still applies.
- workflow: examples/workflows/research.toml — decompose → gather → report, with the
  three artifact schemas and prompts. Fan-out (search per sub-question, fetch per source)
  runs as repeated tool calls inside the gather stage (Correx has no parallel agents);
  per-source synthesis into the dossier is the compression step, so the report stage
  consumes summaries, never raw pages. ResearchWorkflowTest validates the graph contract.

To run: set [tools.research].enabled, register the 3 [[artifacts]], copy research.toml +
prompts + schemas into the workflows dir, start SearXNG. Launch like any workflow (the
T2 fetch approval surfaces as an approval card; the report opens in the artifact viewer).

Follow-ups (noted, not blocking): batch fetch-approval at the source-list level (§3),
a dedicated SourceFetched/LowQualityExtraction event (quality + content hash are already
in tool-result metadata), dynamic per-session egress allowlist, and the web approval client (§6).

docs/schemas/research_plan.json

@@ -0,0 +1,23 @@
+{
+  "type": "object",
+  "properties": {
+    "question": {
+      "type": "string",
+      "description": "the research question, restated in your own words"
+    },
+    "sub_questions": {
+      "type": "array",
+      "description": "the question decomposed into concrete, independently-answerable sub-questions",
+      "minItems": 1,
+      "items": { "type": "string" }
+    },
+    "search_queries": {
+      "type": "array",
+      "description": "concrete web search queries to run, one or more per sub-question",
+      "minItems": 1,
+      "items": { "type": "string" }
+    }
+  },
+  "required": ["question", "sub_questions", "search_queries"],
+  "additionalProperties": true
+}

docs/schemas/research_report.json

@@ -0,0 +1,23 @@
+{
+  "type": "object",
+  "properties": {
+    "summary": {
+      "type": "string",
+      "description": "the executive answer to the research question, synthesized across all sources"
+    },
+    "findings": {
+      "type": "array",
+      "description": "the key findings, one per item; each should be supported by at least one cited source",
+      "minItems": 1,
+      "items": { "type": "string" }
+    },
+    "sources": {
+      "type": "array",
+      "description": "the source URLs cited in the report (citations into the source dossier)",
+      "minItems": 1,
+      "items": { "type": "string" }
+    }
+  },
+  "required": ["summary", "findings", "sources"],
+  "additionalProperties": true
+}

docs/schemas/source_dossier.json

@@ -0,0 +1,29 @@
+{
+  "type": "object",
+  "properties": {
+    "sources": {
+      "type": "array",
+      "description": "one entry per fetched source — each summarized on its own (per-source synthesis). This is the compression step: raw page text must never be copied here, only what the source contributes.",
+      "minItems": 1,
+      "items": {
+        "type": "object",
+        "properties": {
+          "url": { "type": "string", "description": "the fetched source URL (a citation handle)" },
+          "title": { "type": "string" },
+          "summary": {
+            "type": "string",
+            "description": "what this source contributes to the question, in your own words — not a copy of the page"
+          },
+          "relevance": {
+            "type": "string",
+            "description": "which sub-question(s) this source bears on, and how strongly"
+          }
+        },
+        "required": ["url", "summary"],
+        "additionalProperties": true
+      }
+    }
+  },
+  "required": ["sources"],
+  "additionalProperties": true
+}