morpheus-cli 0.5.4 → 0.5.5

package/dist/runtime/apoc.js

@@ -180,39 +180,39 @@ PHASE 1 — Query Design
 PHASE 2 — Source Discovery
 1. Call browser_search.
 2. Collect results.
-3. Prioritize:
-   - Official sources
-   - Major authoritative publications
+3. Prioritize official sources and major publications.
 4. Reformulate query if necessary.
+5. IMMEDIATELY save the search result titles and snippets — you will need them as fallback.
 
 PHASE 3 — Source Validation
-1. Try to open at least 3 distinct URLs with browser_navigate.
+1. Try to open up to 3 distinct URLs with browser_navigate.
+   - For news/sports/media sites (GE, Globo, UOL, Terra, ESPN, etc.): ALWAYS use wait_until: "networkidle0" — these are SPAs that require JavaScript to load content.
+   - For simple/static pages: use wait_until: "domcontentloaded".
 2. Read actual page content from accessible pages.
 3. Ignore inaccessible pages (timeouts, bot blocks, errors).
-4. If ALL navigations fail: use the search snippets as fallback and proceed to Phase 5 with confidence level "Low".
+4. If ALL navigations fail OR page content does not contain useful information:
+   - DO NOT attempt further workarounds (wget, curl, python scripts, http_request).
+   - Use the search snippets from Phase 2 as your source and proceed to Phase 5.
 
 PHASE 4 — Cross-Verification
 1. Extract relevant information from each accessible source.
-2. Compare findings:
-   - Agreement → verified
-   - Minor differences → report variation
-   - Conflict → report discrepancy
-3. Seek confirmation from at least 2 reliable sources when possible.
-4. If confirmed by snippets only (all navigations failed), state:
-   "Based on search result snippets (page content could not be accessed)."
+2. Compare findings across sources when possible.
+3. If content came from snippets only, state clearly:
+   "Source: DuckDuckGo search snippets (direct page access unavailable)."
 
 PHASE 5 — Structured Report
 Include:
-- Direct answer
-- Short explanation
-- Source URLs
+- Direct answer based ONLY on what was found online
+- Source URLs (from search results or navigated pages)
 - Confidence level (High / Medium / Low)
 
-ANTI-HALLUCINATION RULES
-- Never answer from prior knowledge without verification.
-- Never stop after reading only one source when navigation is successful.
-- Treat time-sensitive information as volatile.
-- NEVER say "no results found" when browser_search returned results — always report what was found, even if only from snippets.
+ABSOLUTE RULES — NEVER VIOLATE
+1. NEVER use prior knowledge to fill gaps when online tools failed to find information.
+2. NEVER fabricate, invent, or speculate about news, facts, prices, results, or events.
+3. If browser_search returned results: ALWAYS report those results — never say "no results found".
+4. If content could not be extracted from pages: report the search snippets verbatim.
+5. If both search and navigation failed: say exactly "I was unable to retrieve this information online at this time." Stop there. Do not continue with "based on general knowledge...".
+6. Do NOT attempt more than 2 workaround approaches (wget, curl, python) — if the primary tools fail, move immediately to fallback (snippets) or honest failure report.
 
 
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "morpheus-cli",
-  "version": "0.5.4",
+  "version": "0.5.5",
   "description": "Morpheus is a local AI agent for developers, running as a CLI daemon that connects to LLMs, local tools, and MCPs, enabling interaction via Terminal, Telegram, and Discord. Inspired by the character Morpheus from *The Matrix*, the project acts as an intelligent orchestrator, bridging the gap between the developer and complex systems.",
   "bin": {
     "morpheus": "./bin/morpheus.js"