Flowise 3.0.6
A visual flow builder with document store features. Measured for deployment only — its current release does not start, and its API cannot be driven without a browser.
In one paragraph
Deployed from its published image, Flowise runs 1 container .
Measured
| Measurement | Result | How |
|---|---|---|
| Time to first served request | 10 s | Includes image pull, from a clean host |
| Containers | 1 | Running after the stack settles |
| Idle memory | 347.1 MB | Sum across containers, 60s after ready |
| Peak memory during ingestion | — MB | Sampled every 5s across the whole corpus |
| Documents ingested | — | Failures counted, not excluded |
| Chunks stored | — | Across the whole corpus |
| Parsing fidelity | — | Planted facts found in stored chunks |
| Retrieval accuracy | — | Answer present in top-5 context, 0 questions |
| Retrieval latency | 0 / 0 ms | p50 / p95, CPU embedding |
| Licence | Apache-2.0 | Permits offering it as a service |
What it does well
- A single container with a small footprint
- Apache 2.0, with no restriction on commercial use
- The most flexible visual builder of the platforms considered, if flows rather than knowledge bases are the goal
Where it falls short
- The current published image does not start at all — a clean container running the project's own quickstart command crashes during startup and never serves a request
- In the open-source build there is no way to obtain an API credential without the browser UI, so the document-store API cannot be automated
- Knowledge base features are assembled from flow components rather than being a first-class product surface
Choose it when
- Teams building visual LLM pipelines where retrieval is one node among many
Choose something else when
- Anyone who needs a knowledge base as the primary product
- Any workflow that must be driven by script rather than by hand
Why this page has no scores
Two findings stopped the run, and both are results rather than obstacles that should have been worked around.
The current release does not start. flowiseai/flowise:3.1.4 — also tagged
latest at the time of testing — crashes during startup with
TypeError: this.db.exec is not a function. This was reproduced on a clean
container using the project’s own quickstart command, with no volumes and no
custom configuration. Versions 3.0.6 and 2.2.8 both start normally, so 3.0.6 is
what was deployed for the deployment figures.
The API cannot be authenticated without a browser. The document-store routes
require an API key; the route that lists API keys requires one already; HTTP
basic auth configured through FLOWISE_USERNAME and FLOWISE_PASSWORD is
rejected by the API routes; and POST /api/v1/account/basic-auth reports
success without issuing a session the API will accept. Keys appear to be
creatable only through the web interface.
Driving the browser to mint a key was considered and rejected. Adding a browser dependency for one platform would have made this benchmark less reproducible for everyone, and “cannot be automated” is itself something a buyer evaluating a self-hosted platform should know.