Public Roadmap
A transparent roadmap for PSNF-Core: what’s shipped, what’s next, and what we’re building toward.
(Order may evolve based on feedback and benchmarks.)
✅ shipped
🟡 in progress
🔵 next
Now (Free)
Core cognition & inspection ✅ shipped
- Concept graph (vocab + links)
- Semantic spine (agent / action / patient + roles)
- Short-term memory & episodic memory
- Search: concept / links / facts
- 3D activity map (concept-state)
Controlled learning ✅ shipped
- “Store as Fact” (explicit facts)
- Separate user memory (can be enabled/disabled per edition)
Next (High impact)
Advanced search (easy, big UX win) 🔵 next
- Search by role: “show actions of X”
- “Why is X connected to Y?” explanation panel
- Path search: X → … → Y (shortest semantic chain)
Typo confirmation (smart UX) 🔵 next
- If lemma not found → suggest closest
- Ask: “Did you mean … ?” before creating new concepts
Knowledge expansion
More fact types (beyond “X is Y”) 🔵 next
- part_of (wheel part_of car)
- used_for (knife used_for cutting)
- causes (smoking causes cancer)
- located_in (heart located_in chest)
- capable_of (bird capable_of fly)
- All facts remain explicit + user-confirmed
Long prompts handling 🔵 next
- Chunking of long text
- Preview “what would be learned”
- Optional: store only selected chunks
Future (killer features)
Pro edition – User memory portability 🔵 next
- Export user memory to JSON
- Import / merge user memories
- Selective reset (facts vs experiences)
-
Enables long-term personalization and shared knowledge packs
Memory ecosystem 🔵 next
- User-created memory packs
- Educational memory bundles (subjects, courses)
- Community / marketplace-ready memory formats
-
“Knowledge becomes portable, inspectable and remixable”
Summaries from the graph 🔵 next
- Local summaries of a topic using known facts + links
- Exportable “study notes”
Concept maps / mind maps 🔵 next
- Auto-generated concept map from neighborhood + paths
- Interactive: click nodes to expand / inspect
- Export as JSON / image (optional)
Explicit semantic queries 🔵 next
- Query language: FIND part_of(car)
- Constraints: role, weight threshold, time window (episodes)
Personality controls 🔵 next
- Style sliders: concise ↔ verbose, formal ↔ playful
- Reasoning transparency: low ↔ high debug