AI features map
Every AI surface in PathNav and where to find it. PathNav uses AI to augment what you've already entered: nothing is generated without your data; nothing is sent to a third-party model that isn't on the free or fixed-cost chain you opted into.
The features
Cross-data insights
Looks across vision, goals, milestones, and habits together. Surfaces patterns, conflicts, coherence issues, causal hints, opportunities, and overall plan health.
How: Auto-fires on first dashboard open if you haven't generated insights in the last week. Click Refresh to re-run any time.
Daily briefing
4-8 personalized items each morning: news, nudges, reflective questions, blockers, milestone reminders. References your specific goals and target dates.
How: Generated overnight as a background job; on-demand button on the Briefing page if you want to refresh.
Goal suggestions
Breaks your vision into 4-7 dated goals each with 2-5 milestones. Respects your red lines, avoids drains, matches your style preference (conservative/balanced/ambitious).
How: Click Generate goal suggestions on Plan home, or run the Tune-up ritual.
Discovery feed
Curated research, repos, reading, and operator notes pulled from your interests, skills, hobbies, public footprint, and active goals. Not a news feed: items are picked because they could change how you work toward a goal.
How: Background job runs every 20 minutes when you have profile context. Items appear under Discovery feed.
Vision improvement
Returns 2-4 sharper variants of your vision (each emphasizing different pillars or identities) plus an honest 2-3 sentence critique. Click a variant to use it.
How: Edit your vision, click ✨ Improve with AI.
Milestone suggestions
Drafts 3-5 dated milestones for an empty goal. Distributes target dates evenly between today and the goal's target. Skips milestones you've already added.
How: Open a goal that has no milestones, click ✨ Suggest with AI. Each proposal is a click-to-add card.
Focus pick rationale
Picks one task from the deterministic top-8 ranking and explains why in 1-2 sentences tied to your vision and current energy state. Optionally suggests an alternative.
How: Switch Focus to the Suggest variant, click ✨ Why this one?
Pillar classifier
Reads the title and pre-selects which of the seven pillars it belongs to, with a confidence score and short explanation.
How: Type a sidequest/someday/wishlist/drain title, click ✨ Auto next to the pillar chips.
Chat (Guide)
Streamed conversation grounded on your vision/goals/projects. Wraps your data in safe XML so prompt-injection in your own notes can't hijack the assistant.
How: Open the chat dock at the bottom-right of any page, or visit /dashboard/chat.
Provider diagnostics
Per-provider health probe (latency + error mode), every key's snooze status, and the last 50 AI calls with their provider/model/latency.
How: Click Run live probes to test each provider with a 1-token ping. Each card shows tier (FREE / FIXED / PAID / SELF) and a sign-up link if no key is configured.
How the provider chain works
- PathNav rotates through 9 AI providers. The chain is free-first: Groq → Cerebras → Gemini → OpenRouter → Mistral → Cloudflare → SambaNova → Ollama → Anthropic.
- You can configure multiple keys per provider (comma-separated in the env var or with _2 / _3 suffixes). Hit a 429 rate limit on one key, the orchestrator snoozes it and rotates to the next.
- Every AI route validates its output schema. If a provider returns unparseable JSON, the next provider in the chain runs. If they all fail, the route falls back to a deterministic template so you never see a complete blank.
- Pay-per-token (Anthropic) only fires if you explicitly set ANTHROPIC_API_KEY. The free chain runs first by default.
You can see live status (which providers have keys, who's snoozed, latest latencies) on the AI diagnostics page.
Privacy posture
PathNav's stance on your data going to AI providers:
- Your vision, goals, milestones, and habits are sent only to the providers you've configured keys for.
- Free-tier providers' terms of service vary; check each one before pasting data you'd consider sensitive.
- Free tier defaults to Groq + Cerebras + Gemini + Mistral + Cloudflare + SambaNova. None of these train on your data under their stated free-tier terms (verify on each provider's site before relying on this).
- BYO Keys mode (developer plan) lets you point at self-hosted Ollama for fully local inference.