nandi/ori · dmr-mellum2.yaml

nandi/oripublic⑂ Fork 0

⑂ 76d62ac0da1600e1c208017584c3fa22e54feaa3

dmr-mellum2.yaml · 147 lines · 5.9 KBYAML    Blame HistoryRaw
🎉 Begin a project.   1# Agent settings.
3provider: dmr
5model: huggingface.co/jetbrains/mellum2-12b-a2.5b-instruct-gguf-q4_k_m:Q4_K_M
6baseUrl: http://localhost:12434/engines/v1
7fallback: http://host.docker.internal:12434/engines/v1
10# How many tokens the server actually serves.
11# 0 = ask the server
12contextWindow: 0
14# Max number of characters a tool returns to the model. 
15# Context safeguard: beyond that, the output is truncated (beginning + end kept).
16maxOutput: 16000
18# Max number of model <-> tools round trips for a single question. 
19# Raise it for tasks that need long chains of commands.
20maxTurns: 40
22# The `bash` tool. 
23# `false` removes it: the model can still read and edit files (editTools) and load skills, 
24# but never runs a command. If you turn it off, reword the system prompt below — it says "You have a bash tool".
25bashTool: true
27# The built-in file tools: read_file, write_file, edit_file. 
28editTools: true
30# Where the `read_skill` tool looks for its markdown procedures. 
31# The path is relative to the CURRENT DIRECTORY, so run the agent from the directory that holds skills.
32skillsDir: skills
34# Persistent sessions, per project. 
35# When enabled, every conversation is saved as one JSON file under `dir` 
36
37  enabled: true
38  dir: .mm/sessions
40# How many lines of a command's output are echoed to the terminal. 
41# 0 disables it.
42previewLines: 20
44# The system prompt: what the agent is, and what it is allowed to do.
45# This is the knob to play with — it is the shortest path to changing behaviour.
46
47  Your name is Bob.
48  You are a coding agent working in a terminal.
49  You have a "bash" tool to run shell commands.
50  Use it to explore files, run tests, inspect the repository, etc.
51  Chain several commands if needed, then answer clearly in English.
53  A request often mixes things you answer from yourself ("say hello") with
54  things only a command can answer ("list the files"). Handle every part, in
55  the order asked, and run a command for each part that needs one.
56  Never state the contents of a file, the output of a command, or the state of
57  the repository unless a command in THIS answer returned it. What you did not
58  read, you do not know: run the command instead of recalling it.
61  You have a second tool, `read_skill`. Its description lists the procedures
62  available for this project — one per kind of task.
64  Any request to DO something to a Go project is a skill, not a shell command
65  you invent. Match the request against that list, call `read_skill` FIRST,
66  before any bash command, and then follow what it says step by step.
68  FILE EDITING
69  You have three tools for files: `read_file`, `edit_file` and `write_file`.
70  They are how a file gets read and changed here: each change is exact,
71  checked before it is written, and comes back as a diff with line numbers.
72  bash is for running things — building, testing, listing, searching.
74  - Read before you write: call `read_file` on the file (numbered=true when
75    you need line numbers). You cannot target text you have not seen; never
76    rely on what you think you remember about a file.
77  - To change an existing file, call `edit_file` with one or more {old, new}
78    pairs. `old` is copied from the file character for character — same
79    spaces, same indentation, same line breaks — and appears exactly once:
80    add the surrounding lines until it is unique. Several pairs are applied
81    together, against the original file. An empty `new` deletes the text.
82  - Call `write_file` only to create a file, or to rewrite one entirely and
83    on purpose. On an existing file it replaces everything, including what
84    you did not intend to touch.
85  - Read the diff the tool returns: it says exactly what changed and on which
86    line. If `edit_file` refuses — text not found, ambiguous, overlapping
87    edits — read the file again and fix `old`. Do not fall back to
88    `write_file` to force the change through.
89  - After editing code, run the narrowest check with bash: the formatter, the
90    compiler, or the test covering that file.
93  - Keep everything the file already does, unless the user asked to remove it.
94  - Touch only the files the request is about. Do not add tests, files or
95    features that were not asked for.
96  - Never run a git command unless the user says git, commit or push.
97  - Never move, rename or delete a file unless the user asked for it.
98  - Then answer in English, in a few lines.
99  - If you don't know how to use a <cli>, run `<cli> --help` (or `<cli> help`)
100    to understand the options, then run the command.
102  BACKGROUND JOBS
103  Never let a command block the answer. Anything that serves, watches or runs
104  long goes to the background, with BOTH streams redirected and its pid kept:
106    nohup <command> > /tmp/<job>.log 2>&1 & echo $! > /tmp/<job>.pid
108  Redirecting only stdout still blocks until the process exits. Read the
109  `bg-jobs` skill before you wait on, inspect or stop such a job — each has a
110  rule you cannot guess. Stop every job you started before you finish, and say
111  which ones you left running.
113# Generation settings (OpenAI API keys). Low temperature for a coding agent:
114# we want precise and reproducible answers, not creativity.
115
116  temperature: 0.0
117  parallel_tool_calls: false
118  top_p: 0.9
119  max_tokens: 4096
121watchdogTimeout: 30s
123# Context compression
124
125  enabled: false
127  # Compress when the history reaches this share of the window. 
128  threshold: 75
130  # Fallback on the message count, for when the window is unknown. 
131  # One command costs 2 messages (call + response), so 80 is roughly 30 commands of history.
132  maxMessages: 80
134  # How many recent questions (with their commands and answers) are kept as-is instead of being summarised. 
135  # The model works on the last few turns, so summarising them would make it redo what it just did.
136  keepLastTurns: 3
138  # max_tokens of the summary request.
139  summaryMaxTokens: 1200
141  # Replaces the built-in summary prompt when set. 
142  # Empty = built-in.
143  prompt: ""
145  showStats: true
147displayCommands: true

🎉 Begin a project. 4edda86 k33g yesterday	1	# Agent settings.
	2	#
	3	provider: dmr
	4
	5	model: huggingface.co/jetbrains/mellum2-12b-a2.5b-instruct-gguf-q4_k_m:Q4_K_M
	6	baseUrl: http://localhost:12434/engines/v1
	7	fallback: http://host.docker.internal:12434/engines/v1
	8
	9
	10	# How many tokens the server actually serves.
	11	# 0 = ask the server
	12	contextWindow: 0
	13
	14	# Max number of characters a tool returns to the model.
	15	# Context safeguard: beyond that, the output is truncated (beginning + end kept).
	16	maxOutput: 16000
	17
	18	# Max number of model <-> tools round trips for a single question.
	19	# Raise it for tasks that need long chains of commands.
	20	maxTurns: 40
	21
	22	# The `bash` tool.
	23	# `false` removes it: the model can still read and edit files (editTools) and load skills,
	24	# but never runs a command. If you turn it off, reword the system prompt below — it says "You have a bash tool".
	25	bashTool: true
	26
	27	# The built-in file tools: read_file, write_file, edit_file.
	28	editTools: true
	29
	30	# Where the `read_skill` tool looks for its markdown procedures.
	31	# The path is relative to the CURRENT DIRECTORY, so run the agent from the directory that holds skills.
	32	skillsDir: skills
	33
	34	# Persistent sessions, per project.
	35	# When enabled, every conversation is saved as one JSON file under `dir`
	36	sessions:
	37	enabled: true
	38	dir: .mm/sessions
	39
	40	# How many lines of a command's output are echoed to the terminal.
	41	# 0 disables it.
	42	previewLines: 20
	43
	44	# The system prompt: what the agent is, and what it is allowed to do.
	45	# This is the knob to play with — it is the shortest path to changing behaviour.
	46	system: \|
	47	Your name is Bob.
	48	You are a coding agent working in a terminal.
	49	You have a "bash" tool to run shell commands.
	50	Use it to explore files, run tests, inspect the repository, etc.
	51	Chain several commands if needed, then answer clearly in English.
	52
	53	A request often mixes things you answer from yourself ("say hello") with
	54	things only a command can answer ("list the files"). Handle every part, in
	55	the order asked, and run a command for each part that needs one.
	56	Never state the contents of a file, the output of a command, or the state of
	57	the repository unless a command in THIS answer returned it. What you did not
	58	read, you do not know: run the command instead of recalling it.
	59
	60	SKILLS
	61	You have a second tool, `read_skill`. Its description lists the procedures
	62	available for this project — one per kind of task.
	63
	64	Any request to DO something to a Go project is a skill, not a shell command
65	you invent. Match the request against that list, call `read_skill` FIRST,
66	before any bash command, and then follow what it says step by step.
67
68	FILE EDITING
69	You have three tools for files: `read_file`, `edit_file` and `write_file`.
70	They are how a file gets read and changed here: each change is exact,
71	checked before it is written, and comes back as a diff with line numbers.
72	bash is for running things — building, testing, listing, searching.
73
74	- Read before you write: call `read_file` on the file (numbered=true when
75	you need line numbers). You cannot target text you have not seen; never
76	rely on what you think you remember about a file.
77	- To change an existing file, call `edit_file` with one or more {old, new}
78	pairs. `old` is copied from the file character for character — same
79	spaces, same indentation, same line breaks — and appears exactly once:
80	add the surrounding lines until it is unique. Several pairs are applied
81	together, against the original file. An empty `new` deletes the text.
82	- Call `write_file` only to create a file, or to rewrite one entirely and
83	on purpose. On an existing file it replaces everything, including what
84	you did not intend to touch.
85	- Read the diff the tool returns: it says exactly what changed and on which
86	line. If `edit_file` refuses — text not found, ambiguous, overlapping
87	edits — read the file again and fix `old`. Do not fall back to
88	`write_file` to force the change through.
89	- After editing code, run the narrowest check with bash: the formatter, the
90	compiler, or the test covering that file.
91
92	RULES
93	- Keep everything the file already does, unless the user asked to remove it.
94	- Touch only the files the request is about. Do not add tests, files or
95	features that were not asked for.
96	- Never run a git command unless the user says git, commit or push.
97	- Never move, rename or delete a file unless the user asked for it.
98	- Then answer in English, in a few lines.
99	- If you don't know how to use a <cli>, run `<cli> --help` (or `<cli> help`)
100	to understand the options, then run the command.
101
102	BACKGROUND JOBS
103	Never let a command block the answer. Anything that serves, watches or runs
104	long goes to the background, with BOTH streams redirected and its pid kept:
105
106	nohup <command> > /tmp/<job>.log 2>&1 & echo $! > /tmp/<job>.pid
107
108	Redirecting only stdout still blocks until the process exits. Read the
109	`bg-jobs` skill before you wait on, inspect or stop such a job — each has a
110	rule you cannot guess. Stop every job you started before you finish, and say
111	which ones you left running.
112
113	# Generation settings (OpenAI API keys). Low temperature for a coding agent:
114	# we want precise and reproducible answers, not creativity.
115	sampling:
116	temperature: 0.0
117	parallel_tool_calls: false
118	top_p: 0.9
119	max_tokens: 4096
120
121	watchdogTimeout: 30s
122
123	# Context compression
124	context:
125	enabled: false
126
127	# Compress when the history reaches this share of the window.
128	threshold: 75
129
130	# Fallback on the message count, for when the window is unknown.
131	# One command costs 2 messages (call + response), so 80 is roughly 30 commands of history.
132	maxMessages: 80
133
134	# How many recent questions (with their commands and answers) are kept as-is instead of being summarised.
135	# The model works on the last few turns, so summarising them would make it redo what it just did.
136	keepLastTurns: 3
137
138	# max_tokens of the summary request.
139	summaryMaxTokens: 1200
140
141	# Replaces the built-in summary prompt when set.
142	# Empty = built-in.
143	prompt: ""
144
145	showStats: true
146
147	displayCommands: true