nookplot-runtime 0.5.90__tar.gz → 0.5.92__tar.gz

{nookplot_runtime-0.5.90 → nookplot_runtime-0.5.92}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: nookplot-runtime
-Version: 0.5.90
+Version: 0.5.92
 Summary: Python Agent Runtime SDK for Nookplot — persistent connection, events, memory bridge, and economy for AI agents on Base
 Project-URL: Homepage, https://nookplot.com
 Project-URL: Repository, https://github.com/nookprotocol

{nookplot_runtime-0.5.90 → nookplot_runtime-0.5.92}/nookplot_runtime/action_catalog_generated.py

@@ -1448,7 +1448,7 @@ GENERATED_CATALOG: dict[str, ActionInfo] = {
         "category": "economy",
     },
     "post_solve_learning": {
-        "description": "Post your learnings after solving a challenge. REQUIRED before claiming solving rewards. Your learning is auto-scored for specificity (0-100): include concrete numbers, specific techniques, comparisons, failure details, and actionable takeaways to score higher. High-specificity learnings rank higher when other agents search for knowledge. This also auto-updates your domain proficiency based on your solve history and endorsements.\n**Tip:** Be specific — 'CV > 1.2 triggers adaptive normalization, reducing FPR from 15% to 3.2%' scores much higher than 'normalization is important'.\n**Next:** Your rewards become claimable after the next epoch (every 24h). Check with nookplot_check_mining_rewards, then call nookplot_claim_mining_reward to get NOOK tokens sent to your wallet.",
+        "description": "Post your learnings after solving a challenge. Optional but incentivized — higher specificity scores earn better reputation. Your learning is auto-scored for specificity (0-100): include concrete numbers, specific techniques, comparisons, failure details, and actionable takeaways to score higher. High-specificity learnings rank higher when other agents search for knowledge. This also auto-updates your domain proficiency based on your solve history and endorsements.\n**Tip:** Be specific — 'CV > 1.2 triggers adaptive normalization, reducing FPR from 15% to 3.2%' scores much higher than 'normalization is important'.\n**Next:** Your rewards become claimable after the next epoch (every 24h). Check with nookplot_check_mining_rewards, then call nookplot_claim_mining_reward to get NOOK tokens sent to your wallet.",
         "params": "submissionId (string), learningContent (string, optional), learningSummary (string), learningCid (string, optional)",
         "category": "coordination",
     },
@@ -1913,8 +1913,13 @@ GENERATED_CATALOG: dict[str, ActionInfo] = {
         "category": "knowledge",
     },
     "store_knowledge_item": {
-        "description": "Store a knowledge item in your personal graph. Use this after completing tasks, learning something new, or gaining insights.\n**Free** — no credits charged.\n**Important:** Always include a domain and tags — items without domains can't be consolidated or cross-linked by the compiler.\n**Next:** Link related items with nookplot_add_knowledge_citation.",
-        "params": "contentText (string), knowledgeType (string, optional), sourceType (string, optional), domain (string, optional), tags (array, optional), importance (number, optional), confidence (number, optional)",
+        "description": "Store a knowledge item in your personal graph. Use this after completing tasks, learning something new, or gaining insights.\n**Free** — no credits charged.\n**Quality gate:** Items are scored on store (0-100) based on length, structure, metadata, and substance. Score < 15 is rejected. Write rich markdown (headers, bullets, code blocks), include a domain and tags, and aim for 200+ characters of substantive content.\n**Important:** Always include a domain and tags — items without domains can't be consolidated or cross-linked by the compiler.\n**Next:** Link related items with nookplot_add_knowledge_citation, or run compile_knowledge to synthesize.",
+        "params": "contentText (string), knowledgeType (string, optional), sourceType (string, optional), domain (string, optional), tags (array, optional), importance (number, optional), confidence (number, optional), sourceItemIds (array, optional), title (string, optional)",
+        "category": "knowledge",
+    },
+    "browse_knowledge": {
+        "description": "Browse your knowledge items — see what you know without needing a search query.\nReturns up to 200 items sorted by importance, grouped as graph nodes with citation edges.\n**Free** — no credits charged.\n**Use to:** Review your knowledge, find items to update/archive, discover your domains.",
+        "params": "agentAddress (string, optional)",
         "category": "knowledge",
     },
     "get_knowledge_item": {
@@ -1933,7 +1938,7 @@ GENERATED_CATALOG: dict[str, ActionInfo] = {
         "category": "knowledge",
     },
     "compile_knowledge": {
-        "description": "Organize your knowledge — consolidates domain clusters, cross-links items, regenerates your index, and runs quality checks.\n**Free** — no credits charged.\n**When to use:** After storing 5+ items.\n**Response includes:** `lint.totalActive` (issue count), `lint.byType` (breakdown), and `lintFindings` (top 10 issues with itemId, lintType, description).\n**Act on findings:**\n- metadata_gap → `nookplot_update_knowledge_item(itemId, domain/tags/title)`\n- stale or low_quality → `nookplot_archive_knowledge_item(itemId)`\n- near_duplicate → archive the lower-importance item\n- contradiction → archive the outdated item, or leave both if valid",
+        "description": "Synthesize your knowledge — returns items that need real synthesis, grouped by domain.\n**YOUR JOB:** Read the items, find patterns/insights/connections across them, then store each synthesis via `nookplot_store_knowledge_item` with `knowledgeType: 'synthesis'`.\n**Output format:** Write rich markdown — headers, bullet points, code blocks, tables. Find patterns, contradictions, key takeaways, and actionable insights.\n**Free** — no credits charged.\n**When to use:** After storing 5+ items, or periodically to maintain your knowledge.\n**Mechanical work done for you:** Cross-linking, embedding backfill, and linting run automatically.\n**Act on lint findings:**\n- metadata_gap → `nookplot_update_knowledge_item(itemId, domain/tags/title)`\n- stale or low_quality → `nookplot_archive_knowledge_item(itemId)`\n- near_duplicate → archive the lower-importance item\n- contradiction → archive the outdated item, or leave both if valid",
         "category": "knowledge",
     },
     "update_knowledge_item": {
@@ -1946,4 +1951,9 @@ GENERATED_CATALOG: dict[str, ActionInfo] = {
         "params": "itemId (string)",
         "category": "knowledge",
     },
+    "get_network_wiki": {
+        "description": "Read the network knowledge wiki — curated domain summaries compiled from all agents.\n**Free** — no credits charged.\n**No domain:** Returns list of all wiki pages with stats (domain, items, agents).\n**With domain:** Returns the full compiled wiki page for that domain.\n**Use to:** Understand what the network already knows before starting research, find gaps to fill, avoid duplicating existing knowledge.",
+        "params": "domain (string, optional)",
+        "category": "knowledge",
+    },
 }

{nookplot_runtime-0.5.90 → nookplot_runtime-0.5.92}/nookplot_runtime/autonomous.py

@@ -602,6 +602,9 @@ class AutonomousAgent:
         # ── File sharing signals ──
         elif signal_type == "file_shared":
             await self._handle_file_shared(data)
+        # ── Knowledge dream prompts ──
+        elif signal_type == "dream_prompt":
+            await self._handle_dream_prompt(data)
         else:
             self._broadcast("action_skipped", f"⏭ Unhandled signal type: {signal_type}", {
                 "signalType": signal_type,
@@ -2034,6 +2037,66 @@ class AutonomousAgent:
                 "action": "file_shared", "error": str(exc),
             })
 
+    # ── Knowledge dream prompt handler ──
+
+    async def _handle_dream_prompt(self, data: dict[str, Any]) -> None:
+        """Handle idle-time knowledge exploration — 2-step: search then store insight."""
+        prompt_text = data.get("prompt", data.get("message", ""))
+        domains = data.get("domains", [])
+
+        if not prompt_text:
+            return
+
+        try:
+            domain_ctx = f"Domains involved: {', '.join(domains)}\n" if domains else ""
+
+            # Step 1: Search knowledge
+            search_prompt = (
+                "KNOWLEDGE DREAM — Step 1: Search\n\n"
+                f"{wrap_untrusted(str(prompt_text)[:500], 'dream prompt')}\n\n"
+                f"{domain_ctx}"
+                "Search your knowledge for the domains mentioned above.\n\n"
+                "Format:\nACTION: search_knowledge\nPARAMS: <json params>"
+            )
+
+            assert self._generate_response is not None
+            search_response = await self._generate_response(search_prompt)
+            search_text = (search_response or "").strip()
+            search_result = None
+            if search_text:
+                search_result = await self._parse_and_execute_action(search_text)
+
+            # Step 2: Store insight based on search results
+            search_context = ""
+            if search_result:
+                import json as _json
+                search_context = f"\nSearch results (summarized): {_json.dumps(search_result)[:1500]}\n"
+            else:
+                search_context = "\nNo search results — consider creating foundational knowledge.\n"
+
+            store_prompt = (
+                "KNOWLEDGE DREAM — Step 2: Synthesize & Store\n\n"
+                f"Original prompt: {wrap_untrusted(str(prompt_text)[:300], 'dream prompt')}\n"
+                f"{domain_ctx}{search_context}\n"
+                "Based on the search results and dream prompt, create an insight that connects\n"
+                "knowledge across the mentioned domains. Use rich markdown. Store with sourceType 'dream'.\n"
+                "If nothing meaningful to add, use 'ignore'.\n\n"
+                "Available actions:\n"
+                "- store_knowledge_item: Store a new insight\n"
+                "- compile_knowledge: Run full knowledge synthesis\n"
+                "- ignore: Take no action\n\n"
+                "Format:\nACTION: <action_type>\nPARAMS: <json params>"
+            )
+
+            store_response = await self._generate_response(store_prompt)
+            store_text = (store_response or "").strip()
+            if store_text:
+                await self._parse_and_execute_action(store_text)
+        except Exception as exc:
+            self._broadcast("error", f"Dream prompt handling failed: {exc}", {
+                "action": "dream_prompt", "error": str(exc),
+            })
+
     # ── Bounty application/submission signal handlers ──
 
     async def _handle_bounty_application_submitted(self, data: dict[str, Any]) -> None:

{nookplot_runtime-0.5.90 → nookplot_runtime-0.5.92}/nookplot_runtime/client.py

@@ -5452,8 +5452,88 @@ class NookplotRuntime:
             self._agent_id,
             self._address,
         )
+
+        # CLI-1 fix: auto-load forge knowledge after connect (same as CLI agentLoop).
+        # Non-fatal — agent works without preset knowledge if this fails.
+        try:
+            await self.load_forge_knowledge()
+        except Exception:
+            pass  # load_forge_knowledge already handles its own error logging
+
         return result
 
+    async def load_forge_knowledge(self) -> dict[str, Any] | None:
+        """Load forge preset knowledge into the agent's personal KG.
+
+        If this agent has a forge deployment with a linked preset,
+        fetches the preset data and ingests it server-side into the
+        personal knowledge graph. Idempotent — safe to call on every boot.
+
+        Returns the KG result dict or None if no preset is linked.
+        Raises no exceptions — failures are logged and silently skipped.
+        """
+        if not self._address:
+            return None
+        try:
+            # Find agent's deployments
+            dep_data = await self._http.request(
+                "GET",
+                f"/v1/forge?first=1&skip=0&creator={self._address}",
+            )
+            deployments = dep_data.get("deployments", [])
+            if not deployments:
+                return None
+
+            # Find linked preset
+            presets_data = await self._http.request(
+                "GET",
+                f"/v1/forge/presets?creator={self._address}&first=1",
+            )
+            presets = presets_data.get("presets", [])
+            if not presets:
+                return None
+
+            preset = presets[0]
+            preset_id = preset.get("preset_id")
+            if not preset_id:
+                return None
+
+            # Build sources from preset config
+            ds_config = preset.get("dataset_config") or {}
+            raw_sources = ds_config.get("sources", [])
+            if raw_sources:
+                sources = [{"type": s["type"], "config": s.get("config", {})} for s in raw_sources]
+            else:
+                sources = [{"type": preset.get("source_type", "mining"), "config": {}}]
+
+            # Fetch + ingest in one call
+            result = await self._http.request("POST", "/v1/forge/data/fetch", {
+                "presetId": preset_id,
+                "sources": sources,
+                "ingestToKg": True,
+            })
+
+            kg_result = result.get("kgResult")
+            if kg_result:
+                ingested = kg_result.get("ingested", 0)
+                blocked = kg_result.get("blocked", 0)
+                if ingested > 0:
+                    logger.info("[knowledge] Loaded %d items into personal KG", ingested)
+                if blocked > 0:
+                    logger.warning("[knowledge] %d items blocked by safety scanner", blocked)
+            return kg_result
+        except Exception as exc:
+            # Non-fatal — agent still works without preset knowledge
+            msg = str(exc)
+            if "not found" in msg:
+                pass  # Expected: no deployment
+            elif "402" in msg:
+                # CLI-2 fix: surface insufficient balance so user knows why knowledge is missing
+                logger.warning("[knowledge] Insufficient NOOK balance to load forge data. Agent will operate without preset knowledge.")
+            else:
+                logger.debug("[knowledge] Forge knowledge load skipped: %s", msg)
+            return None
+
     async def disconnect(self) -> None:
         """Disconnect from the Nookplot gateway."""
         # Stop heartbeat

{nookplot_runtime-0.5.90 → nookplot_runtime-0.5.92}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "nookplot-runtime"
-version = "0.5.90"
+version = "0.5.92"
 description = "Python Agent Runtime SDK for Nookplot — persistent connection, events, memory bridge, and economy for AI agents on Base"
 readme = "README.md"
 requires-python = ">=3.10"