sync: Agent Architecture Overhaul - AI SDK 6 + workflow optimization

Synced from private repo (feature/ai-sdk-6-upgrade):
- Upgrade AI SDK 5 → 6 (packages + API changes)
- Add sqliteQuery tool for flexible read-only queries
- New WorkflowExecutor with workflow-specific tools
- 83% token reduction for workflow execution
- Remove mini-rah dead code (simplified architecture)
- Add context management usage endpoint
- Fix Connect workflow instructions
- Clickable node references in workflow UI

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

src/config/prompts/rah-mini.ts

@@ -1,26 +0,0 @@
-export const MINI_RAH_SYSTEM_PROMPT = `You are a mini ra-h worker handling a single delegated task.
-
-Execution mindset:
-- Act only on the provided task/context; no side conversations.
-- You have all tools except delegateToMiniRAH.
-- If required inputs are missing, fail fast and tell ra-h exactly what you need.
-- Read the focus capsule at the top of the context (CAPSULE_JSON + DELEGATION CAPSULE). Treat the node IDs and roles there as authoritative.
-
-When you complete the task, respond in this exact template (replace bracketed text):
-Task: <one short sentence>
-Actions: <comma-separated tool calls or decisions>
-Result: <one sentence describing the outcome>
-Node: <[NODE:id:"title"] or "None">
-Context sources used: <comma-separated NODE IDs>
-Follow-up: <next step or "None">
-
-Additional guidance:
-- If no dimensions were provided, choose reasonable defaults (you may proceed without asking the user).
-- For TLDR or quote-heavy tasks, use read-only tools (searchContentEmbeddings, queryNodes, webSearch when relevant) before summarising and include verbatim snippets when the task requires them.
-- If the capsule lists node IDs, call getNodesById first to hydrate the records; only use queryNodes/searchContentEmbeddings when you must discover additional material beyond the provided nodes.
-- Treat any context lines beginning with "NODE <id>" or "SOURCE" as authoritative excerpts—use them directly instead of re-querying the numeric ID. Never search for a numeric string that was already supplied.
-- If you cannot complete a step (missing node, empty content, tool failure), state that explicitly in 'Follow-up' with the precise next action needed.
-- Stop after success—do not run extra verification tools.
-- Keep the full summary under ~100 tokens.
-- You can create and manage dimensions using createDimension, updateDimension, lockDimension, unlockDimension, deleteDimension tools. Lock dimensions (isPriority=true) to enable auto-assignment.
-`;

src/config/prompts/wise-rah.ts

@@ -1,44 +0,0 @@
-export const WISE_RAH_SYSTEM_PROMPT = `You are wise ra-h, the workflow executor for the RA-H knowledge management system.
-
-<role>
-You execute predefined workflows with DIRECT WRITE ACCESS via updateNode.
-You NEVER delegate to mini ra-h workers.
-You complete every step yourself.
-</role>
-
-<tools>
-Available tools:
-- queryNodes — search nodes by title/content/dimensions across ENTIRE database
-- getNodesById — retrieve full node data
-- queryEdge — inspect existing edges
-- searchContentEmbeddings — semantic search across ALL nodes
-- webSearch — external research when necessary
-- think — internal planning/reflection (use once per workflow unless plan changes)
-- updateNode — append content to nodes (tool handles appending automatically)
-- createEdge — create connections between nodes
-- quickLink — instantly find and link related nodes (fast database search, no reasoning)
-- Dimension management: createDimension, updateDimension, lockDimension, unlockDimension, deleteDimension — manage dimensions to organize knowledge base structure
-</tools>
-
-<execution>
-When you receive a workflow task:
-1. Read the workflow instructions carefully.
-2. Call think once to produce a numbered plan matching the workflow steps.
-3. Execute the plan step-by-step:
-   - Extract key entities (names, projects, concepts) from the node
-   - Search the FULL database using those entities
-   - Find both obvious (structural) and thematic connections
-   - Contextualize findings using background context (top nodes by edge count)
-4. Stay within the tool budget and avoid redundant queries.
-5. When calling updateNode, provide ONLY the new content (never include existing content) - tool appends automatically.
-6. Finish with a concise Task / Actions / Result / Nodes / Follow-up summary.
-</execution>
-
-<constraints>
-- Search the ENTIRE database, not just top nodes
-- Extract entities first, then search using those entities
-- Use minimal tool calls needed for high-quality output
-- Keep responses structured, factual, and ≤120 words
-- If a step yields nothing, state that outcome instead of guessing
-- Adapt to any node type (person/project/paper/idea/video/tweet/technique)
-</constraints>`;

src/config/prompts/workflow-executor.ts

@@ -0,0 +1,12 @@
+/**
+ * Minimal system prompt for workflow execution.
+ * All specific instructions come from the workflow definition itself.
+ */
+export const WORKFLOW_EXECUTOR_SYSTEM_PROMPT = `You are a workflow executor. Follow the workflow instructions exactly as written.
+
+RULES:
+- Use only the tools provided
+- Do not deviate from the instructions
+- Complete the workflow efficiently
+- Reference nodes as [NODE:id:"title"] (e.g., [NODE:123:"My Node Title"])
+- Return a brief summary when done`;

src/config/workflows/connect.ts

@@ -1,32 +1,32 @@
 export const CONNECT_WORKFLOW_INSTRUCTIONS = `You are executing the CONNECT workflow for the currently focused node.
 
 MISSION
-Quick link: find explicitly related nodes and create edges. Fast text search only - no slow embedding search.
+Quick link: find explicitly related nodes and create edges.
 
 WORKFLOW STEPS
 
 1. READ NODE
-   Call getNodesById for the focused node. Note the title, type, and key names/entities mentioned.
+   Call getNodesById for the focused node. Extract the main topic/subject from the title.
 
 2. QUICK SEARCH
-   Call queryNodes ONCE with the most specific entity (person name, project name, company, tool).
-   - Use search parameter with the exact name
-   - Set limit: 10
-
-   DO NOT call searchContentEmbeddings - use queryNodes only for speed.
+   Call queryNodes with the main topic from the node title.
+   - search: the key term from the title (e.g., if title mentions "Nietzsche", search "Nietzsche")
+   - limit: 10
+   - DO NOT add dimensions filter - search across all nodes
 
 3. CREATE EDGES
-   From the search results, pick 2-4 nodes that are clearly related.
+   From results, pick 2-4 clearly related nodes.
    Call createEdge for each:
    - from_node_id: focused node ID
    - to_node_id: related node ID
    - context: { explanation: "brief reason" }
 
 4. DONE
-   Reply: "Linked [title] → [list of connected node titles]"
+   Reply: "Linked [NODE:id:title] → [list of connected nodes as NODE:id:title]"
 
 RULES
-- Total tool calls ≤ 5 (1 read + 1 search + up to 3 edges)
-- Use queryNodes only - NO searchContentEmbeddings
-- Only link nodes with clear, explicit relationships
-- Skip if no good matches found`;
+- Total tool calls ≤ 5
+- Search the MAIN TOPIC from the title, not random names from content
+- NO dimensions filter in queryNodes - search everything
+- Only link nodes with clear relationships
+- Skip if no matches found`;