Boson AI · Modal overflow

Provision & e2e verification evidence

← Back to plan & tracker

wan raw-server provisioning + 74-API verification (2026-07-02, via live URL)

URL: https://bosonai--wan-web-stream-serve.modal.run

Provision time (measured through the URL, no fabrication)

COLD PROVISION: 131s warm req 1: 200 0.458487s warm req 2: 200 0.396619s warm req 3: 200 0.378007s (container-internal boot to /health+openapi: 144s — wan_boot_measure.py)

All 74 endpoints — HTTP code from curl (SERVED = any HTTP status)

first sweep (sequential):

200 POST /avatar/debug/sessions 000 DELETE /avatar/debug/sessions/{session_id} 000 POST /avatar/debug/sessions/{session_id}/finalize 000 POST /avatar/debug/sessions/{session_id}/step 400 POST /avatar/realtime/reference 000 POST /avatar/realtime/sessions 000 DELETE /avatar/realtime/sessions/{session_id} 422 POST /avatar/realtime/sessions/{session_id}/audio 000 POST /avatar/realtime/sessions/{session_id}/start 000 POST /avatar/video/sessions 404 GET /avatar/video/sessions/{session_id} 404 DELETE /avatar/video/sessions/{session_id} 422 POST /avatar/video/sessions/{session_id}/audio 404 POST /avatar/video/sessions/{session_id}/audio/end 200 GET /avatar/video/stream/replica/status 200 POST /avatar/video/stream/sessions 404 DELETE /avatar/video/stream/sessions/{session_id} 422 POST /avatar/video/stream/sessions/{session_id}/audio 404 POST /avatar/video/stream/sessions/{session_id}/audio.bin 404 POST /avatar/video/stream/sessions/{session_id}/audio/end 404 GET /avatar/video/stream/sessions/{session_id}/file.mp4 404 GET /avatar/video/stream/sessions/{session_id}/fragments 404 GET /avatar/video/stream/sessions/{session_id}/init.fmp4 404 GET /avatar/video/stream/sessions/{session_id}/profile 404 GET /avatar/video/stream/sessions/{session_id}/status 000 POST /get_weights_checksum 200 GET /health 200 GET /health_generate 200 GET /model_info 200 GET /models 200 GET /server_info 400 POST /update_weights_from_disk 422 POST /v1/images/edits 422 POST /v1/images/generations 404 GET /v1/images/{image_id}/content 000 GET /v1/list_loras 408 POST /v1/merge_lora_weights 422 POST /v1/meshes 200 GET /v1/meshes 404 GET /v1/meshes/{mesh_id} 404 DELETE /v1/meshes/{mesh_id} 404 GET /v1/meshes/{mesh_id}/content 200 GET /v1/model_info 200 GET /v1/models 404 GET /v1/models/{model} 422 POST /v1/set_lora 200 POST /v1/unmerge_lora_weights 400 POST /v1/videos 200 GET /v1/videos 404 GET /v1/videos/{video_id} 404 DELETE /v1/videos/{video_id} 404 GET /v1/videos/{video_id}/content 422 POST /vertex_generate 200 POST /wan_s2v/offline/sessions 404 GET /wan_s2v/offline/sessions/{session_id} 404 DELETE /wan_s2v/offline/sessions/{session_id} 422 POST /wan_s2v/offline/sessions/{session_id}/audio 404 POST /wan_s2v/offline/sessions/{session_id}/audio/end 404 GET /wan_s2v/offline/sessions/{session_id}/chunks 000 POST /wan_s2v/session 000 POST /wan_s2v/session/v3 400 POST /wan_s2v/session/v3/encode_ref 200 DELETE /wan_s2v/session/{session_id} 422 POST /wan_s2v/session/{session_id}/audio 200 POST /wan_s2v/session/{session_id}/finalize 200 GET /wan_s2v/session/{session_id}/pull 200 POST /wan_s2v/session/{session_id}/start 200 POST /wan_s2v/session/{session_id}/step 200 DELETE /wan_s2v/session/{session_id}/v3 200 GET /wan_s2v/stream/replica/status 200 POST /wan_s2v/stream/sessions 404 DELETE /wan_s2v/stream/sessions/{session_id} 422 POST /wan_s2v/stream/sessions/{session_id}/audio 404 POST /wan_s2v/stream/sessions/{session_id}/audio.bin 404 POST /wan_s2v/stream/sessions/{session_id}/audio/end 404 GET /wan_s2v/stream/sessions/{session_id}/file.mp4 404 GET /wan_s2v/stream/sessions/{session_id}/fragments 404 GET /wan_s2v/stream/sessions/{session_id}/init.fmp4 404 GET /wan_s2v/stream/sessions/{session_id}/profile 404 GET /wan_s2v/stream/sessions/{session_id}/status

11 that timed out (000) on the sweep, re-tested in isolation -> all served:

 200 0.910149s  POST /wan_s2v/session
 200 1.364613s  POST /wan_s2v/session/v3
 200 0.658652s  POST /avatar/video/sessions
 200 0.780698s  POST /avatar/realtime/sessions
200 10.613302s  POST /get_weights_checksum
 200 0.653236s  GET /v1/list_loras
 200 1.356873s  POST /avatar/debug/sessions/test/step
 200 2.718683s  POST /avatar/debug/sessions/test/finalize
 200 1.001387s  DELETE /avatar/debug/sessions/test
 400 1.051662s  POST /avatar/realtime/reference

200 POST /avatar/realtime/sessions/{id}/start (21.5s) 200 DELETE /avatar/realtime/sessions/{id} (7.6s)


Avatar endpoint re-validation — 2026-07-06 (3 days unattended, hardened config)

URL: https://bosonai--avatar.modal.run (official higgs-avatar v1.0.0, warmup-gated routing, watchdog, 5-min liveness probe, min_containers=1)

Check Result
/health 200 in 0.46 s
replica/status accepts_new_session:true, worker idle
browser player 200
full e2e render (milestones/v1_e2e_client.py) PASS — init.fmp4 3.8 s, fragments streamed, file.mp4 618,973 B valid, 9.1 s total