Uses text-to-image generative AI to create an ephemeral "photo feed" of the Bluesky firehose.
See it at imagehose.net
Behavior is controlled via environment variables, see Usage below for details.
$ npm start
Or with docker:
$ docker build -t imagehose .
$ docker run ... imagehose
Any not marked as required have reasonable defaults.
| Variable name | Description | Req'd? |
|---|---|---|
IMAGEHOSE_EXPECT_IMG_GEN_TIME_S |
The rough expected image generation time: it's not tracked and accounted for explicitly, instead this simple fudge is used. Will be subtracted from IMAGEHOSE_ARTIFICAL_DELAY_S |
No |
IMAGEHOSE_PLOT_DATA_RETAIN_X_MORE_SAMPLES |
Retain this many times more samples for statistics than are retained for posts themselves (IMAGEHOSE_POST_COUNT) |
No |
IMAGEHOSE_PLOT_X_PRCNT_MIN_AVERAGES |
Percentage of total samples retained (IMAGEHOSE_PLOT_DATA_RETAIN_X_MORE_SAMPLES * IMAGEHOSE_POST_COUNT) to use in the windowed average calculations. |
No |
IMAGEHOSE_POST_COUNT |
Number of posts to retain before deleting the oldest. | No |
IMAGEHOSE_ARTIFICAL_DELAY_S |
Time, in seconds, to wait between each generation. Candidate posts are collected during this wait period. | No |
IMAGEHOSE_WAIT_FOR_ATLEAST |
Number of posts to wait for as candidates before first generation, after the process is started fresh. | No |
IMAGEHOSE_ADDL_FILTERS_JSON |
A JSON list of additional filter words to append to the profanity list generated from @dsojevic/profanity-list |
No |
IMAGEHOSE_IMG_GEN_URL_PREFIX |
The full URL prefix, including the token & prompt=, of an img-gen-cf instance. |
YES |
IMAGEHOSE_MODEL_URL |
The text-to-image model's URL | YES |
IMAGEHOSE_MODEL_NAME |
The text-to-image model's name | YES |
IMAGEHOSE_PAGE_AUTO_REFRESH_S |
How often the generated pages are set to auto-refresh | No |
IMAGEHOSE_WSS_SERVER_PORT |
Websocket server port | No |
IMAGEHOSE_FQDN |
Host fully-qualified domain name with protocol | YES |