1# SPDX-FileCopyrightText: Amolith <amolith@secluded.site>
 2#
 3# SPDX-License-Identifier: CC0-1.0
 4
 5# Example custom models configuration for rumilo
 6# Copy relevant sections to your $XDG_CONFIG_HOME/rumilo/config.toml
 7
 8[defaults]
 9model = "anthropic:claude-sonnet-4-20250514"
10cleanup = true
11# kagi_session_token = "your-kagi-token"
12# tabstack_api_key = "your-tabstack-key"
13
14[web]
15model = "anthropic:claude-sonnet-4-20250514"
16# system_prompt_path = "~/.config/rumilo/web-prompt.txt"
17
18[repo]
19model = "anthropic:claude-sonnet-4-20250514"
20default_depth = 1
21blob_limit = "5m"
22# system_prompt_path = "~/.config/rumilo/repo-prompt.txt"
23
24# Custom models allow you to use any OpenAI-compatible endpoint
25# Reference them with: rumilo web "query" --model custom:ollama
26
27[custom_models.ollama]
28# Ollama running locally
29provider = "ollama"
30api = "openai-completions"
31base_url = "http://localhost:11434/v1"
32id = "ollama/llama3.2"
33name = "Llama 3.2 (Ollama)"
34reasoning = false
35input = ["text"]
36cost = { input = 0, output = 0 }
37context_window = 128000
38max_tokens = 4096
39
40[custom_models.vllm]
41# vLLM inference server
42provider = "vllm"
43api = "openai-completions"
44base_url = "http://localhost:8000/v1"
45id = "meta-llama/Llama-3.3-70B-Instruct"
46name = "Llama 3.3 70B (vLLM)"
47reasoning = false
48input = ["text"]
49cost = { input = 0, output = 0 }
50context_window = 131072
51max_tokens = 8192
52
53[custom_models.groq]
54# Groq (built-in to pi-ai, but shown here as example)
55provider = "groq"
56api = "openai-completions"
57base_url = "https://api.groq.com/openai/v1"
58id = "groq/llama-3.3-70b-versatile"
59name = "Llama 3.3 70B Versatile (Groq)"
60reasoning = false
61input = ["text"]
62cost = { input = 0.59, output = 0.79 }
63context_window = 131072
64max_tokens = 8192
65
66[custom_models.openrouter]
67# OpenRouter API (model aggregation)
68provider = "openrouter"
69api = "openai-completions"
70base_url = "https://openrouter.ai/api/v1"
71id = "openrouter/zai/glm-4.5v"
72name = "GLM-4.5V (OpenRouter)"
73reasoning = true
74input = ["text", "image"]
75cost = { input = 0.5, output = 1.5 }
76context_window = 128000
77max_tokens = 4096