-
Notifications
You must be signed in to change notification settings - Fork 415
Expand file tree
/
Copy pathcapability-overrides.yaml
More file actions
45 lines (41 loc) · 1.66 KB
/
Copy pathcapability-overrides.yaml
File metadata and controls
45 lines (41 loc) · 1.66 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
# Demonstrates the `capabilities` override for models the models.dev catalogue
# does not describe correctly. Custom OpenAI-compatible providers and local
# models are not (always) in models.dev, so docker-agent cannot auto-detect
# whether they accept image/PDF attachments and conservatively falls back to
# text-only, silently dropping attachments (issue #2741).
#
# Declaring `capabilities` makes the model's attachment support authoritative
# and skips the models.dev lookup entirely. The flags must match what the
# endpoint actually accepts; claiming an unsupported modality leads to a
# provider-side API error.
providers:
# A local Ollama endpoint (OpenAI-compatible). models.dev has no "ollama"
# provider entry and never catalogues arbitrary local tags such as "llava".
ollama:
base_url: http://localhost:11434/v1
# A self-contained custom OpenAI-compatible gateway. Its provider name is
# user-defined, so models.dev knows nothing about the models it serves.
vision-proxy:
api_type: openai_chatcompletions
base_url: https://llm.internal.example.com/v1
token_key: MY_PROXY_API_KEY
models:
# Local vision model: the override is the only way to keep image attachments.
llava-local:
provider: ollama
model: llava
capabilities:
image: true
pdf: false
# Custom gateway serving a vision- and PDF-capable model.
proxy-vision:
provider: vision-proxy
model: gpt-4o
capabilities:
image: true
pdf: true
agents:
root:
model: llava-local
description: Assistant on a local Ollama vision model with declared image support
instruction: You are a helpful assistant. Describe any images the user attaches.