Repository navigation
Expand file tree
/
Copy pathdocker-compose.yaml
More file actions
521 lines (508 loc) · 19.6 KB
/
Copy pathdocker-compose.yaml
File metadata and controls
521 lines (508 loc) · 19.6 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
services:
# wait when resolve freeze + OOM https://github.com/antirez/ds4/issues/705
ds4:
image: ds4:spark-full
build:
context: .
dockerfile: Dockerfile.ds4.spark
volumes:
- ${HOME}/src/github.com/antirez/ds4/gguf:/app/gguf
- ${HOME}/src/github.com/antirez/ds4/server-kv:/app/server-kv
command: /app/ds4-server --cuda --host 0.0.0.0 --prefill-chunk 2048 --port 8200 --threads 20 --kv-disk-dir /app/server-kv --kv-disk-space-mb 8192 --dspark --model /app/gguf/DeepSeek-V4-Flash-Layers37-42Q4KExperts-OtherExpertLayersIQ2XXSGateUp-Q2KDown-AProjQ8-SExpQ8-OutQ8-chat-v2-imatrix-fixed-0731.gguf --mtp /app/gguf/DeepSeek-V4-Flash-DSpark-support-0731.gguf
networks:
- spark-network
ports:
- "127.0.0.1:8200:8200"
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: 1
capabilities: [ gpu ]
healthcheck:
test: [ "CMD", "curl", "-f", "http://localhost:8200/health" ]
interval: 60s
timeout: 10s
retries: 3
start_period: 10s
ds4-sparkinfer:
# wait when merge https://github.com/0xSero/deepseek-v4-flash-0731-spark-sparkinfer/pull/2
# image: ghcr.io/0xsero/deepseek-v4-flash-0731-spark-sparkinfer@sha256:2e077489a83a0360952828051fe7f7a32c1801e5ce8436d85f7267583d614ff4
image: ds4-sparkinfer:spark-full
build:
context: ${HOME}/src/github.com/0xSero/deepseek-v4-flash-0731-spark-sparkinfer
dockerfile: Dockerfile
ipc: host
shm_size: 16gb
environment:
HF_TOKEN: ${HF_TOKEN:-}
MODEL_REPO: 0xSero/deepseek-v4-flash-0731-spark
MODEL_REVISION: 22f28d32b9b29b4352eaa380ff8c2c170b2847ab
MAX_MODEL_LEN: "262144"
MAX_NUM_SEQS: "4"
MAX_NUM_BATCHED_TOKENS: "8224"
MODE: dspark
DSPARK_TOKENS: "5"
DSPARK_CAPACITY: "0"
DSPARK_DYNAMIC_DRAFT_DEPTH: "0"
DSPARK_DYNAMIC_DRAFT_DEPTH_WINDOW: "8"
DSPARK_DRAFT_EXPERTS: "64"
DSPARK_STRUCTURED_EXPERTS_PER_CATEGORY: "32"
VLLM_USE_B12X_WO_PROJECTION: "1"
GPU_MEMORY_UTILIZATION: "0.9465"
VERIFY_MODEL_CHECKSUMS: "1"
volumes:
- ${HOME}/src/github.com/0xSero/deepseek-v4-flash-0731-spark-sparkinfer/data:/models
- ${HOME}/src/github.com/0xSero/deepseek-v4-flash-0731-spark-sparkinfer/cache:/cache
networks:
- spark-network
ports:
- "127.0.0.1:8000:8000"
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: 1
capabilities: [ gpu ]
healthcheck:
test: [ "CMD", "curl", "-f", "--connect-timeout 2" ,"--max-time 5", "http://localhost:8000/health" ]
interval: 60s
timeout: 10s
retries: 3
start_period: 10s
max-inference:
image: modular/max-nvidia-full:latest
volumes:
- ${HOME}/.cache/huggingface:/root/.cache/huggingface:rw
- ${HOME}/.cache/max_cache:/opt/venv/share/max/.max_cache:rw
command:
- --port
- "8100"
- --model
- Qwen/Qwen3-Coder-30B-A3B-Instruct-FP8
- --trust-remote-code
ports:
- "127.0.0.1:8100:8100"
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: 1
capabilities: [ gpu ]
networks:
- spark-network
environment:
- HF_TOKEN=${HF_TOKEN:-}
healthcheck:
test: [ "CMD", "curl", "-f", "http://localhost:8100/health" ]
interval: 60s
timeout: 10s
retries: 3
start_period: 10s
sglang:
image: lmsysorg/sglang:dev-cu13
healthcheck:
test: [ "CMD", "curl", "-f", "http://localhost:30000/health" ]
interval: 60s
timeout: 10s
retries: 5
start_period: 180s
volumes:
- ${HOME}/.cache/huggingface:/models/huggingface:ro
shm_size: '32gb'
networks:
- spark-network
environment:
- HF_HOME=/models/huggingface
- HF_HUB_OFFLINE=1
- HF_TOKEN=${HF_TOKEN:-}
ports:
- "127.0.0.1:30000:30000"
command: >
sglang serve
--port 30000
--host 0.0.0.0
--model-path ${SGLANG_MODEL:-Qwen/Qwen3.5-27B-FP8}
--context-length 261375
--kv-cache-dtype fp8_e4m3
--tp 1
--reasoning-parser qwen3
--tool-call-parser qwen3_coder
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: 1
capabilities: [ gpu ]
vllm:
# image: vllm/vllm-openai:cu130-nightly-aarch64
# image: ghcr.io/aeon-7/vllm-aeon-ultimate-dflash:qwen36-v3
# look build-vllm-cuda13-docker.sh
image: ${VLLM_IMAGE:-vllm:spark-full}
# build:
# context: .
# dockerfile: Dockerfile.vllm.spark
volumes:
- ${HOME}/.cache/huggingface:/models/huggingface
# qwen38-flash-next: vLLM caches + packed PLE table (~27GB, ~/.cache/vllm/ple_cache).
# The old named volume kept spark-full compile/ccache state across images - swap
# this line back together with VLLM_IMAGE=vllm:spark-full in .env
- ${HOME}/.cache/vllm:/root/.cache/vllm
# - vllm-cache:/root/.cache/
- ./docker/vllm-configs:/etc/vllm-configs:ro
# qwen38-flash-next runtime patches (generated from the image by
# ~/MiaAI-Lab/Qwen3.8-Flash-Next-Single-DGX-Spark/files/patch_*.py - edit generators,
# not these copies). qwen38-flash-next PROFILE ONLY - re-enable together with
# VLLM_IMAGE=vllm/vllm-openai:qwen38-flash-next in .env. They overwrite shared vllm
# modules (modelopt.py carries MarlinNvFp4LinearKernel!) and break other images:
# the flint image dies in weight load (shape assert) with modelopt_patched mounted.
# Verified 2026-09-07: all still required - image tag qwen38-flash-next not rebuilt
# since 2026-08-26 and nothing merged upstream yet:
# - PLE CPU offload + NVFP4 dispatch: open PR vllm-project/vllm#53899
# https://github.com/vllm-project/vllm/pull/53899 (conflicts, awaiting review)
# - GB10/TP=1 offload deadlock (host-side handshake replaces CUDA stream mem ops):
# open bug vllm-project/vllm#53960 https://github.com/vllm-project/vllm/issues/53960
# - QSA fp8_e4m3 KV cache: open RFC vllm-project/vllm#54426
# https://github.com/vllm-project/vllm/issues/54426 (reimplemented from
# https://github.com/lancelind/qwen3.8-Flash-DGX, Apache-2.0)
# - ModelOpt MXFP8 small-shape BF16 fallback (FlashInfer needs N,K>=128) and the
# MTP draft-vocab patch (inert without VLLM_MTP_DRAFT_VOCAB): no upstream PR yet
# When #53899 merges and a new image ships: regenerate patches (or drop the mounts
# if the release notes say GB10 offload landed), see files/patch_*.py in the MiaAI repo.
# - ./docker/vllm-patches/qwen38-flash-next/ple_layer_patched.py:/usr/local/lib/python3.12/dist-packages/vllm/models/qwen3_8_flash_next/nvidia/ple_layer.py:ro
# - ./docker/vllm-patches/qwen38-flash-next/modelopt_patched.py:/usr/local/lib/python3.12/dist-packages/vllm/model_executor/layers/quantization/modelopt.py:ro
# - ./docker/vllm-patches/qwen38-flash-next/qsa_ops_patched.py:/usr/local/lib/python3.12/dist-packages/vllm/models/qwen3_8_flash_next/nvidia/ops/qsa.py:ro
# - ./docker/vllm-patches/qwen38-flash-next/qsa_nvidia_patched.py:/usr/local/lib/python3.12/dist-packages/vllm/models/qwen3_8_flash_next/nvidia/qsa.py:ro
# - ./docker/vllm-patches/qwen38-flash-next/mtp_patched.py:/usr/local/lib/python3.12/dist-packages/vllm/models/qwen3_8_flash_next/nvidia/mtp.py:ro
# - ./docker/vllm-patches/qwen38-flash-next/ple_offload_layer.py:/usr/local/lib/python3.12/dist-packages/vllm/model_executor/layers/ple_offload_layer.py:ro
# - ./docker/vllm-patches/qwen38-flash-next/ple_offload_connector.py:/usr/local/lib/python3.12/dist-packages/vllm/v1/ple_offload/connector.py:ro
# - ./docker/vllm-patches/qwen38-flash-next/ple_offload_worker.py:/usr/local/lib/python3.12/dist-packages/vllm/v1/ple_offload/worker.py:ro
# - ./docker/vllm-patches/qwen38-flash-next/ple_offload_protocol.py:/usr/local/lib/python3.12/dist-packages/vllm/v1/ple_offload/protocol.py:ro
# Recursive Flint (vllm PR #62938): vllm:flint image was built from PR commit db02bc9f;
# six later PR commits (826fd3f..02f263a) gate the custom SM100 kernels off on sm_120/121
# and fix the stock-layers fallback (weight load shape assert dies on GB10 without them).
# Files extracted from pr-62938 branch head - drop these mounts when a rebuilt flint
# image ships (or the PR merges and lands in a release/nightly)
- ./docker/vllm-patches/flint/model.py:/usr/local/lib/python3.12/dist-packages/vllm/models/flint/model.py:ro
- ./docker/vllm-patches/flint/configs.py:/usr/local/lib/python3.12/dist-packages/vllm/models/flint/configs.py:ro
- ./docker/vllm-patches/flint/gguf_nvfp4.py:/usr/local/lib/python3.12/dist-packages/vllm/models/flint/gguf_nvfp4.py:ro
- ./docker/vllm-patches/flint/__init__.py:/usr/local/lib/python3.12/dist-packages/vllm/models/flint/__init__.py:ro
- ./docker/vllm-patches/flint/ops/__init__.py:/usr/local/lib/python3.12/dist-packages/vllm/models/flint/ops/__init__.py:ro
- ./docker/vllm-patches/flint/ops/cute_primitives.py:/usr/local/lib/python3.12/dist-packages/vllm/models/flint/ops/cute_primitives.py:ro
- ./docker/vllm-patches/flint/ops/decode_attention.py:/usr/local/lib/python3.12/dist-packages/vllm/models/flint/ops/decode_attention.py:ro
- ./docker/vllm-patches/flint/ops/linear.py:/usr/local/lib/python3.12/dist-packages/vllm/models/flint/ops/linear.py:ro
- ./docker/vllm-patches/flint/ops/moe.py:/usr/local/lib/python3.12/dist-packages/vllm/models/flint/ops/moe.py:ro
- ./docker/vllm-patches/flint/ops/moe_cute.py:/usr/local/lib/python3.12/dist-packages/vllm/models/flint/ops/moe_cute.py:ro
- ./docker/vllm-patches/flint/ops/moe_gemv.py:/usr/local/lib/python3.12/dist-packages/vllm/models/flint/ops/moe_gemv.py:ro
- ./docker/vllm-patches/flint/ops/norm.py:/usr/local/lib/python3.12/dist-packages/vllm/models/flint/ops/norm.py:ro
- ./docker/vllm-patches/flint/ops/nvfp4_gemv.py:/usr/local/lib/python3.12/dist-packages/vllm/models/flint/ops/nvfp4_gemv.py:ro
- ./docker/vllm-patches/flint/ops/qk_norm_rope.py:/usr/local/lib/python3.12/dist-packages/vllm/models/flint/ops/qk_norm_rope.py:ro
- ./docker/vllm-patches/flint/ops/qkv_fused.py:/usr/local/lib/python3.12/dist-packages/vllm/models/flint/ops/qkv_fused.py:ro
# - ./docker/chat-template-qwen3.jinja:/etc/vllm/chat-template.jinja:ro
# - ./docker/safetensors_coding_generation.json:/models/huggingface/hub/models--AxionML--Qwen3.5-35B-A3B-NVFP4/snapshots/75f5a9d934e5528d372916d8df7376ec52f6801a/generation_config.json:ro
environment:
- HF_HOME=/models/huggingface
- HF_HUB_OFFLINE=1
- HF_TOKEN=${HF_TOKEN:-}
# ─── Core vLLM ───
- VLLM_ALLOW_LONG_MAX_MODEL_LEN=1
# ─── GB10 / sm_121a runtime ───
- TORCH_CUDA_ARCH_LIST=12.1a
- PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True
- NVIDIA_FORWARD_COMPAT=1
- NVIDIA_DISABLE_REQUIRE=1
# ─── FP4 path selection (qwen38-era knobs, NOT for the flint image) ───
# - VLLM_USE_FLASHINFER_MOE_FP4=1
# - VLLM_TEST_FORCE_FP8_MARLIN=0
# - VLLM_USE_FLASHINFER_SAMPLER=1
# - VLLM_NVFP4_GEMM_BACKEND=flashinfer-cutlass
# - ENABLE_NVFP4_SM120=1
# - TORCH_MATMUL_PRECISION=high
# - CUTE_DSL_ARCH=sm_121a
# unified memory
- VLLM_WORKER_MULTIPROC_METHOD=spawn
# VLLM_USE_DEEP_GEMM is env-only (no CLI flag, so it can't live in vllm-configs/*.yaml);
# set VLLM_USE_DEEP_GEMM=0 in .env for FP8 models that crash with
# deepgemm "Unknown SF transformation" (/vllm/.deps/deepgemm-src/csrc/apis/layout.hpp:60)
- VLLM_USE_DEEP_GEMM=${VLLM_USE_DEEP_GEMM:-1}
# flashinfer runtime compilation
- MAX_JOBS=4
ports:
- "127.0.0.1:30001:30001"
# qwen38-flash-next PLE offload worker uses POSIX shm (MiaAI recipe runs --ipc host);
# leftover /dev/shm segments leak until reboot when the container is SIGKILLed
ipc: host
ulimits:
memlock:
soft: -1
hard: -1
stack: 67108864
cap_add:
- SYS_NICE
- SYS_PTRACE
# vLLM start really slow =(
healthcheck:
test: [ "CMD", "curl", "-f", "http://localhost:30001/health" ]
interval: 30s
timeout: 10s
retries: 100
start_period: 100s
networks:
- spark-network
# per-model configs live in docker/vllm-configs/*.yaml, selected via VLLM_CONFIG in .env
# vllm/vllm-openai:qwen38-flash-next ships ENTRYPOINT ["vllm","serve"], spark-full uses
# nvidia_entrypoint.sh - pin the entrypoint so the command below works for both
# (vllm:flint tag = Dockerfile.vllm.spark flint target, keeps the same entrypoint)
entrypoint: ["vllm", "serve"]
command: >
--config /etc/vllm-configs/${VLLM_CONFIG:-qwen3.6-35b-a3b-fp8.yaml}
--host 0.0.0.0
--port 30001
# GB10: CUDA unified memory is host-backed and charged to this cgroup.
# Without a limit a runaway load (e.g. runai_streamer staging 71GB+) eats all RAM + swap,
# and the node OOM killer never fires because swap reclaim "makes progress" -> machine thrashes.
# mem_limit + memswap_limit (equal) make the kernel OOM-kill only this container, early, no swap.
mem_limit: 112g
memswap_limit: 112g
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: 1
capabilities: [ gpu ]
paroquant:
# image: ghcr.io/z-lab/paroquant:serve
image: paroquant:spark-full
build:
context: .
dockerfile: Dockerfile.paroquant.spark
ports:
- "127.0.0.1:30002:30002"
volumes:
- ${HOME}/.cache/huggingface:/models/huggingface
environment:
- HF_HOME=/models/huggingface
- HF_TOKEN=${HF_TOKEN:-}
- HF_HUB_OFFLINE=1
- TRANSFORMERS_OFFLINE=1
- VLLM_CACHE_ROOT=/models/huggingface
- VLLM_WORKER_MULTIPROC_METHOD=spawn
entrypoint:
- /bin/bash
- -xc
command: >
"
python3 -m paroquant.cli.serve
--model ${PAROQUANT_MODEL:-z-lab/Qwen3.6-27B-PARO}
--host 0.0.0.0
--port 30002
--max-model-len 261375
--max-num-batched-tokens 8192
--enable-chunked-prefill
--max-num-seqs 2
--kv-cache-dtype fp8
--gpu-memory-utilization 0.72
--trust-remote-code
--enable-prefix-caching
--enable-auto-tool-choice
--tool-call-parser gemma4
--reasoning-parser gemma4
"
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: 1
capabilities: [ gpu ]
atlas:
image: avarok/atlas-gb10:latest
ipc: host
volumes:
- ~/.cache/huggingface:/root/.cache/huggingface
environment:
- RUST_LOG=info
command: >
serve ${ATLAS_MODEL:-Qwen/Qwen3.6-35B-A3B-FP8}
--bind 0.0.0.0
--port 30003
--max-seq-len 261375
--max-batch-size 4
--kv-cache-dtype ${ATLAS_KV_CACHE_TYPE:-fp8}
--kv-high-precision-layers auto
--gpu-memory-utilization ${ATLAS_MEMORY_USAGE:-0.9}
--scheduling-policy slai
--enable-prefix-caching
--max-prefill-tokens 8192
--speculative
--num-drafts 2
--tool-call-parser qwen3_coder
ports:
- "127.0.0.1:30003:30003"
ulimits:
memlock:
soft: -1
hard: -1
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: 1
capabilities: [ gpu ]
networks:
- spark-network
healthcheck:
test: ['CMD', 'bash', '-c', 'exec 3<>/dev/tcp/localhost/30003; printf "GET /health HTTP/1.0\r\n\r\n" >&3; read -r _ code _ <&3; [ "$$code" = 200 ]']
interval: 30s
timeout: 10s
retries: 3
start_period: 10s
llama.cpp:
build:
context: .
dockerfile: Dockerfile.llama.cpp.spark
image: "llama.cpp:spark-full"
# image: ghcr.io/ggml-org/llama.cpp:full-cuda13
volumes:
- ${HOME}/.cache/huggingface:/models/huggingface:ro
- ${HOME}/.cache/modelscope:/models/modelscope:ro
- ./llama.cpp.models.ini:/etc/llama.cpp/config.ini
entrypoint: /app/llama-server
command:
- --models-max
- "2"
- --models-preset
- /etc/llama.cpp/config.ini
- --host
- 0.0.0.0
- --port
- "8090"
- "--metrics"
ports:
- "127.0.0.1:8090:8090"
ulimits:
memlock:
soft: -1
hard: -1
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: 1
capabilities: [ gpu ]
networks:
- spark-network
healthcheck:
test: [ "CMD", "curl", "-f", "http://localhost:8090/health" ]
interval: 60s
timeout: 10s
retries: 3
start_period: 10s
bonsai:
# Ternary Bonsai 2 27B (1.72 bpw ternary g128 + Hadamard activation runtime)
# needs PrismML-Eng/llama.cpp fork - stock llama.cpp rejects PQ2_0/PTQ1_0 (ggml types 142/143),
# no upstream support yet: https://github.com/ggml-org/llama.cpp/issues/29058
# image built by build-llama.cpp-prism-docker.sh (fork releases have no aarch64 CUDA binaries)
image: llama.cpp:prism-full
volumes:
- ${HOME}/.cache/huggingface:/models/huggingface:ro
- ./llama.cpp.models.ini:/etc/llama.cpp/config.ini
entrypoint: /app/llama-server
command:
- --models-max
- "2"
- --models-preset
- /etc/llama.cpp/config.ini
- --host
- 0.0.0.0
- --port
- "8091"
- "--metrics"
ports:
- "127.0.0.1:8091:8091"
ulimits:
memlock:
soft: -1
hard: -1
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: 1
capabilities: [ gpu ]
networks:
- spark-network
healthcheck:
test: [ "CMD", "curl", "-f", "http://localhost:8091/health" ]
interval: 60s
timeout: 10s
retries: 3
start_period: 10s
vibe-kanban:
build:
context: .
dockerfile: Dockerfile.vibe-kanban
image: vibe-kanban:spark
volumes:
- ${HOME}:/root/
- /tmp:/tmp
ports:
- "127.0.0.1:${VIBE_KANBAN_PORT:-8888}:${VIBE_KANBAN_PORT:-8888}"
networks:
- spark-network
environment:
- HOME=/root
- AGENT_INFERENCE_SERVER=${AGENT_INFERENCE_SERVER:-max-inference}
- AGENT_MAIN_MODEL=${AGENT_MAIN_MODEL:-noctrex/MiniMax-M2-REAP-139B-A10B-MXFP4_MOE-GGUF}
- AGENT_BACKGROUND_MODEL=${AGENT_BACKGROUND_MODEL:-noctrex/MiniMax-M2-REAP-139B-A10B-MXFP4_MOE-GGUF}
healthcheck:
test: [ "CMD", "curl", "-f", "http://localhost:8888/health" ]
interval: 60s
timeout: 10s
retries: 3
start_period: 10s
seer-marketing:
image: ghcr.io/mx-seer/seer:latest
volumes:
- ./docker/seer-config.yml:/app/config.yaml
ports:
- "127.0.0.1:${SEER_PORT:-8889}:${SEER_PORT:-8889}"
networks:
- spark-network
aider-desk:
image: ghcr.io/hotovo/aider-desk:latest
ports:
- "127.0.0.1:${AIDER_DESK_PORT:-24337}:${AIDER_DESK_PORT:-24337}"
volumes:
- aider-desk-data:/app/data/
- ${HOME}/src/:/src/
- ${HOME}/.aider-desk/:/root/.aider-desk/
networks:
- spark-network
b4:
image: lavrushin/b4:latest
container_name: b4
network_mode: host
cap_add:
- NET_ADMIN
- NET_RAW
- SYS_MODULE
volumes:
- ./docker/b4:/etc/b4
command: ["--config", "/etc/b4/b4.json"]
restart: unless-stopped
# docker network create spark-network
networks:
spark-network:
external: true
volumes:
aider-desk-data:
vllm-cache: