PathMap Alchemy

Autonomous mode: seed with two concepts as usual. After each build, Alchemy picks one random, not-yet-run follow-up pair from every suggestion collected so far and runs it, then repeats. It never repeats a pair, and stops only when no new pairs remain or you press Cancel. Every run is saved to your library.

Question / cue (optional)

0:00

Library

Saved automatically in this browser. Click a molecule or MeSH tag to see every run that shares it.

Import trace / JSON / dataset

Graph

→

Dig Deeper with Google Dig Deeper with ChatGPT

PathMap Alchemy v1.0.1 · Built by Joshua Dungan, inventor of pathmap.org and founder of Artificial General Intelligence LLC, using free-tier Claude and Google AI tools, and made with open-weight model use in mind.

Credits: PathMap (Apache 2.0) · JavaScript, HTML5 and SVG (one static file, no backend) · browser Web Storage (localStorage) for your library · PubMed via NCBI E-utilities · NCBI Taxonomy · PubChem PUG REST and PUG View · Gemma open-weight models and the Google Gemini API · Ollama · LM Studio · OpenRouter · Neo4j Cypher (export format).

API key Model Get a free key

Endpoint

Model name

Start Ollama with OLLAMA_ORIGINS=* (or enable CORS in LM Studio) so the browser may call it.

API key ModelGet a key

Limit Gemma models to a random sample of 16 abstracts per prompt (keeps small context windows from overflowing). Turn this off for localhost or paid APIs with large context.

Your API key is never stored. It is held in memory for this tab only and is sent only to the provider's own API URL. Reloading the page clears it.

Share this result?

When this run finishes you can contribute the result to the global research library at phosita.tech/intake.php, or keep it private on this device.

Shared: your concepts and cue exactly as typed (do not put private notes in them), the final write-up, verified quotes with PMIDs, molecule data, the model name, the app version, and the full run trace (prompts, model responses, PubMed/PubChem requests, build metadata and build log). Removed before sending: API keys, local or private network endpoints, your browser details, and email addresses. Never shared: other library runs. Accepted results may be published openly with a permanent identifier and included in open datasets for research and model training. Nothing is sent unless you choose Share. You can reset this choice in About.

Remember my choice

How to use PathMap Alchemy

About PathMap Alchemy

PathMap Alchemy runs client-side using open-weight models (Gemma) and local inference (Ollama). By eliminating server-side compute overhead, it demonstrates how open innovation enables zero-cost, privacy-first micro-SaaS tools for independent researchers.

Architecture

One static HTML file, no backend. Model calls go from your browser to Gemini, your localhost, or OpenRouter. Literature comes from PubMed E-Utilities; molecule data from PubChem and NCBI Taxonomy. Non-secret settings and your library live in this browser's localStorage. API keys are never stored: they stay in memory for this tab only and are sent only to the provider's own API URL. Sharing a result with the global library is optional and asked each time unless you choose to remember it. Based on PathMap (Apache 2.0, Joshua Dungan / Artificial General Intelligence LLC).