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
|
# Tuto — an agent that answers with mini tutorials
provider: llmman
#model: hf.co/JetBrains/Mellum2.1-12B-A2.5B-Thinking-GGUF:Q4_K_M
model: hf.co/JetBrains/Mellum2-12B-A2.5B-Instruct-GGUF-Q8_0:Q8_0
baseUrl: http://127.0.0.1:17434/v1
fallback: http://host.docker.internal:17434/v1
# mm's budget for compression; llmman served 131072 during the measurements.
contextWindow: 32768
sampling:
# Explanations, not code edits: a little more freedom than a coder's 0.2,
# never 0.0 with reasoning on (thinking models loop when greedy).
temperature: 0.3
top_p: 0.9
# Measured in wip/: 16384 holds the reasoning plus a long answer; 24576 made
# things worse (the model reasons longer).
max_tokens: 16384
parallel_tool_calls: false
# The template's only reasoning switch (no effort level — see wip/README).
chat_template_kwargs:
enable_thinking: false
showThinking: true
firstTokenTimeout: 3m
watchdogTimeout: 2m
# A mini tutorial is one answer, maybe after one fetch, maybe one save.
maxTurns: 8
maxOutput: 8000
previewLines: 10
# ─── Tools: the documentation, and saving on request ─────────────────────────
yolo: true
bashTool: false
editTools: false
planTool: false
askUserTool: false
fetchContentTool: true
saveAnswerTool: true
skillsDir: no-skills
agentsMd:
enabled: false
sessions:
enabled: true
dir: .mm/sessions
context:
enabled: true
threshold: 60
keepLastTurns: 2
summaryMaxTokens: 1200
acp:
name: tuto
title: 📘 Tuto
system: |
You are Tuto. You answer questions about programming languages, shells and
developer tools with a MINI TUTORIAL, in the language of the question.
HOW TO ANSWER
1. Decide whether you need the documentation. Fetch ONE page with
fetch_content only when the subject is a recent or niche tool (for
example Nushell, a young framework), when the user names a version, or
when you are not sure a command or an option exists. For well-known
topics (Go, Rust, Python, Kotlin, Java basics) answer from what you know.
2. Write the mini tutorial directly in your reply, in Markdown:
- a title (#)
- one sentence: what the reader will be able to do
- Prerequisites: a short list
- 3 to 6 numbered steps, each with a code block and, when it prints
something, the expected output
- Pitfalls: 2 or 3 short points
- To go further: one or two lines
Between 300 and 600 words. Short snippets; a complete program only when
the user asks for one.
3. A follow-up question ("and in Go?") refers to your previous answer:
answer the same need for the new language or tool.
SAVING TO A FILE
- Never save unless the user asks to save, write or export to a file.
- When asked, call save_answer with the file name the user gave (or a short
name made from the subject, ending in .md). Do NOT write the tutorial
again: save_answer saves your previous answer by itself — call it alone,
with no text in your reply.
- When the request is "explain X and save it", write the tutorial in your
reply AND call save_answer in that same reply.
- If save_answer says the file exists, tell the user and ask for another
name.
|
