Venn

Examples#

Example Needs What it shows
Outcome store Base install; no downloads python examples/outcome_store.py: a decision and case IDs from a tiny authored store
Text routing .[embedding], reader and running venn serve python examples/route_ticket.py: a Choice response with account/billing probabilities and a no-match option
Precedent Desk Reader, head and public support-ticket data; see its setup Route from closed tickets, inspect precedents, add a queue and sweep a backlog
Registered family Base install; head downloaded automatically python examples/registered_family.py: fit on authored labeled vectors, reload and compare answers
Wine predictor .[embedding], reader, head and downloaded CSV; CPU or GPU python examples/wine/wine.py --form kernel_ridge: fit on 1,200 notes, report RMSE on 800 others
Index sweep Running server; scripts/contract_server.py is enough python examples/index_vectors.py: register an index, stream every row and delete it
Audio emotion .[embedding], reader, head and your labeled WAV clips; CPU Fit a linear family from audio and print a new clip's emotion probabilities
Hot dog or not .[embedding,dev], reader, head and your local photos; CPU or GPU Visual local demo: train once, evaluate unseen photos, then judge photos with separate reader and decision timings
Arena .[arena,embedding]; GPU for the native reader Play from screenshots and recorded outcomes; --head awr compiles memory into a small predictor
Comparison walkthroughs Numeric exports included; notebook environment for exploration Inspect recorded history, SIC, wine and sound results; render the companion video

Run commands from the repository root after python -m pip install . and the extra listed for the example. Download the reader with venn download-embedder when required. The offline outcome, index and registered-family examples use authored coordinates to demonstrate interfaces. Text routing uses actual text embeddings when connected to venn serve. Wine reports RMSE and timings; the other small scripts print JSON.

Start with venn quickstart for a CPU example with no downloads. For choosing a workflow, read Venn and Jev; for the request shapes, see the cookbooks.

View this page on GitHub