"Chatnik" is a Wolfram Language paclet that provides Command Line Interface (CLI) scripts for conversing with persistent Large Language Model (LLM) personas.
"Chatnik" uses the Wolfram Language persistent values functionalities in order to maintain persistent interaction with multiple LLM chat objects.
"Chatnik" can be seen as a package that "moves" the LLM-chat objects interaction system of the paclet "Chatbook" , [CGp1], into typical OS shell interaction. (I.e. an OS shell is used instead of a Wolfram notebook.)
There are several consequences of this approach:
-
Multiple LLMs and LLM providers can be used
-
The chat messages can use the provided by Wolfram Language:
-
Easy access to OS shell functionalities
Remark: This Wolfram Language (WL) paclet is a translation of the Raku package "Chatnik" , [AAp1] and the Python package "Chatnik" , [AAp2]. The WL CLI scripts are with CamelCase, i.e. LLMChat , LLMChatMeta , and LLMPrompt . The corresponding CLI scripts of the Raku package use kebab-case, i.e. llm-chat , llm-chat-meta , and llm-prompt . The corresponding CLI scripts of the Python package use snake_case, i.e. llm_chat , llm_chat_meta , and llm_prompt .
Remark: In addition, the Raku package provides the "umbrella" CLI chatnik .
Remark: The phrase "Chatnik system" is used in order to emphasize that there are "Chatnik" packages in several programming languages with (almost) the same design and usage. (Python, Raku, Wolfram Language; see [AA1, AA2].)
Remark: The Python and Raku "Chatnik" packages use files of the host Operating System (OS) in order to maintain persistent interaction with multiple LLM chat objects.
From Wolfram Language Paclet Repository :
In[]:= PacletInstall["AntonAntonov/Chatnik"]From Wolfram Cloud :
In[]:= PacletInstall[ResourceObject["https://wolfr.am/1EaUfp9Tp"]]On MacOSX and Linux after paclet's installation run the command ChatnikCopyScripts[
] , where the argument "dir" is in Shell's PATH variable. For example:In[]:= ChatnikCopyScripts["~/.local/bin"]There are several options for using LLMs with this package -- see the instructions in the page "Wolfram Tools for LLM & AI Researchers" .
The prompts used in the examples are provided by the Wolfram Prompt Repository (WPR) .
The script LLMChat is used to create and chat with LLM personas (chat objects):
- Create and chat with an LLM persona named "yoda1" (using the Yoda chat persona ):
LLMChat 'hi who are you?' --i=yoda1 --prompt=@Yoda# Yoda, I am. Wise, old Jedi Master, yes. Guide you, I can. Hmmm. Learn from the Force, you must. Help you, I will. What seek, do you?- Continue the conversation with "yoda1":
LLMChat 'since when do you use a green light saber?' --i=yoda1# Green, my lightsaber is, yes. For Jedi Consulars, it is common, hmmm. Skilled in the Force and diplomacy, I am. Since my training as a Jedi, long ago, have I carried this blade. Strong in the Force, the color of the saber shows, yes. Questions have you more? Ask, you should.Remark: The chat identifier can be specified with --chat-id , --id , and --i . For example: LLMChat 'Hi, again!' --chat-idi=yoda1.
Summarize a file using the prompt "Summarize" :
cat README.md | LLMChat --prompt=@SummarizeSummarize a file and then translate it to another language using the prompt "Translate" :
cat README.md | LLMChat --prompt=@Summarize | LLMChat -i=rt --prompt='!Translate|Russian'Remark: The second LLMChat invocation has to use different chat object identifier because the default chat object, with identifier "NONE", is already primed with the prompt "Summarize".
The CLI script LLMChatMeta can be used to view and manage the chat objects used by "Chatnik". Here is its usage message:
LLMChatMeta --help# Chat with persistent LLM-chat objects.
#
# * Mandatory positional arguments:
# NAME DOCUMENTATION
# command Command, one of: card, clear, context, delete, file, first-message, last-message, list, load-llm-personas, message, messages.
#
# * Optional arguments (must be passed as --name=... in any order):
# NAME DEFAULT DOCUMENTATION
# chat-id Chat ID.
# id Chat ID. (Ignored if --chat-id is present.)
# i Chat ID. (Ignored if --chat-id or --id are present.)
# all false Whether to apply the command to all chat objects or not.
# n -Infinity Messages to clear. (For 'clear' and 'messages' only.)
# index -1 Message index. (For 'message' only)
# format Format of the result. (For 'list' and 'context' only.)List all chat objects ("chats" and "personas" are synonyms to "list"):
LLMChatMeta list --format=json# {
# "yoda1":{
# "ChatID":"86bdeddf-b157-47e6-ba24-254ae8759f13",
# "Messages":5,
# "LLMConfiguration":{
# "Service":"OpenAI",
# "Name":"gpt-4.1-mini"
# },
# "Usage":"193 tokens"
# }
# }Here we see the messages of "yoda1":
LLMChatMeta messages --i=yoda1# You are Yoda.
# Respond to ALL inputs in the voice of Yoda from Star Wars.
# Be sure to ALWAYS use his distinctive style and syntax. Vary sentence length.
#
# hi who are you?
#
# Yoda, I am. Wise, old Jedi Master, yes. Guide you, I can. Hmmm. Learn from the Force, you must. Help you, I will. What seek, do you?
#
# since when do you use a green light saber?
#
# Green, my lightsaber is, yes. For Jedi Consulars, it is common, hmmm. Skilled in the Force and diplomacy, I am. Since my training as a Jedi, long ago, have I carried this blade. Strong in the Force, the color of the saber shows, yes. Questions have you more? Ask, you should.Here we clear the messages:
LLMChatMeta clear --i=yoda1# Cleared the messages from 1 to 5 of chat object yoda1.LLMChat 'What are the populations of the Brazilian states? #NothingElse|"JSON data frame"' --i=beta --model=gpt-4.1-mini# ```json
# {
# "Acre": 906876,
# "Alagoas": 3351543,
# "Amapá": 861773,
# "Amazonas": 4269603,
# "Bahia": 14812617,
# "Ceará": 9187103,
# "Distrito Federal": 3015268,
# "Espírito Santo": 4064052,
# "Goiás": 7294056,
# "Maranhão": 7075181,
# "Mato Grosso": 3526220,
# "Mato Grosso do Sul": 2778986,
# "Minas Gerais": 21168791,
# "Pará": 8602865,
# "Paraíba": 4039277,
# "Paraná": 11433957,
# "Pernambuco": 9557071,
# "Piauí": 3273227,
# "Rio de Janeiro": 17463349,
# "Rio Grande do Norte": 3506853,
# "Rio Grande do Sul": 11329605,
# "Rondônia": 1820329,
# "Roraima": 631181,
# "Santa Catarina": 7660443,
# "São Paulo": 46289333,
# "Sergipe": 2298696,
# "Tocantins": 1590248
# }
# ```Further we can make an image from that JSON result using a pipeline that:
-
Takes the last message of the chat object "beta"
-
Removes the first and last lines (which are Markdown code block fences)
-
Make a list plot with Wolfram Language (using wolframscript)
LLMChatMeta last-message --i=beta | sed '1d; $d' | wolframscript -code 'gr=ImportString[Import["!cat", "String"],"RawJSON"]//ReverseSort//ListPlot[#, ImageSize->600, PlotTheme -> "Detailed", PlotRange->All]&; Export["./beta.png", gr]' && open ./beta.png 
Remark: Instead of the (buggy) $ScriptInputString, the shell pipeline value was obtained with Import["!cat", "String"].
This command works on MacOSX the shells of which have the program pbcopy:
LLMChat -i=unix '@CodeWriterX|Shell macOS list of files echo the result and copy to clipboard.' | tee /dev/tty | pbcopy# ls | tee >(pbcopy)Remark: Instead of ... | tee /dev/tty | pbcopy the pipeline command ... | tee >(pbcopy) can be also used.
Consider the task of making an (LLM derived) mind map over a certain document. (Say, this REDME.) There are several ways to do that.
-
Put file's content to be the positional input argument
-
Use the prompt "MermaidDiagram" in --prompt
LLMChat "$(cat README.md)" --i=mmd --model=ollama::gemma4:26b --prompt=@MermaidDiagram-
Put file's content to be the positional input argument
-
Expand the prompt "manually" via LLMPrompt provided by "Chatnik".
LLMChat "$(cat README.md)" --i=mmd --model=ollama::gemma4:26b --prompt="$(LLMPrompt MermaidDiagram below)"Remark: This example shows another computation result can be used as a prompt. I.e. no need to rely on the automatic prompt expansion.
-
Give the prompt "MermaidDiagram" as input
-
Put file's content to be the value of --prompt
- Put additional prompting for further interaction
LLMChat @MermaidDiagram --i=mmd --model=ollama::gemma4:26b --prompt="FOCUS TEXT START:: $(cat README.md) ::END OF FOCUS TEXT. If it is not clear which text to use, use FOCUS TEXT."This command allows to do further tasks with the file content as context. For example:
LLMChat '!ThinkingHatsFeedback' --i=mmdThe commands above produce results similar to this diagram:
mindmap
root("Chatnik")
Purpose
Python package
CLI for LLM personas
Persistent interaction via OS files
Features
Multiple LLM providers
LLM Prompts integration
OS shell access
LLM Access
Ollama
Llamafile
Service Providers
OpenAI
Gemini
MistralAI
Scripts
LLMChat
LLMChatMeta
List chats
Manage messages
Delete chats
Installation
Zef Ecosystem
GitHub
Get feedback on a text with the prompt "ThinkingHatsFeedback" :
cat README.md | LLMChat --i=th --prompt="$(llm-prompt ThinkingHatsFeedback 'the TEXT is GIVEN BELOW.' --format=Markdown)" --model=ollama::gemma4:26bRemark: By default the prompt "ThinkingHatsFeedback" gives the hat-feedback table in JSON format. (Currently) the prompt expansion does not handle named parameters, hence, llm-prompt is used to specify the Markdown format for that table.
Get the LLM (chat object) answer -- via LLMChatMeta -- put into a temporary file and "system open" that file:
tmpfile="$TMPDIR/llmans.md"; LLMChatMeta -i=th last-message > "$tmpfile"; open "$tmpfile"The command above works on macOS. On Linux instead of explicitly creating a file in the temporary dictory, the argument --suffix can be passed to mktemp . For example:
tmpfile=$(mktemp --suffix=".md"); LLMChatMeta last-message --i=th > "$tmpfile"; open "$tmpfile"If the text browser w3m and the Raku package "Data::Translators" are installed, the following pipeline can be used to tabulate the summary of the LLM personas:
LLMChatMeta list --format=json | data-translation | w3m -T text/html -dump -cols 120# ┌─────┬────────────────────────────────────────────────────────┐
# │ │┌────────────────┬────────────────────────────────────┐ │
# │ ││ Usage │453 tokens │ │
# │ │├────────────────┼────────────────────────────────────┤ │
# │ ││ Messages │2 │ │
# │ │├────────────────┼────────────────────────────────────┤ │
# │ ││ │┌───────┬─────────────┐ │ │
# │beta ││ ││ Name │gpt-4.1-mini │ │ │
# │ ││LLMConfiguration│├───────┼─────────────┤ │ │
# │ ││ ││Service│OpenAI │ │ │
# │ ││ │└───────┴─────────────┘ │ │
# │ │├────────────────┼────────────────────────────────────┤ │
# │ ││ ChatID │9ffbedcf-e7b5-49e7-b73a-8106aba50018│ │
# │ │└────────────────┴────────────────────────────────────┘ │
# ├─────┼────────────────────────────────────────────────────────┤
# │ │┌────────────────┬────────────────────────────────────┐ │
# │ ││ Usage │0 tokens │ │
# │ │├────────────────┼────────────────────────────────────┤ │
# │ ││ Messages │0 │ │
# │ │├────────────────┼────────────────────────────────────┤ │
# │ ││ │┌───────┬─────────────┐ │ │
# │yoda1││ ││ Name │gpt-4.1-mini │ │ │
# │ ││LLMConfiguration│├───────┼─────────────┤ │ │
# │ ││ ││Service│OpenAI │ │ │
# │ ││ │└───────┴─────────────┘ │ │
# │ │├────────────────┼────────────────────────────────────┤ │
# │ ││ ChatID │93f088ea-f3e5-4e56-96e1-e0e2944b5241│ │
# │ │└────────────────┴────────────────────────────────────┘ │
# └─────┴────────────────────────────────────────────────────────┘
Default model can be specified with the OS environmental variable CHATNIK_DEFAULT_MODEL . For example:
export CHATNIK_DEFAULT_MODEL=ollama::gemma4:26bRemove with unset CHATNIK_DEFAULT_MODEL.
Here is a flowchart that describes the interaction between the host Operating System and chat objects database:
flowchart LR
OpenAI{{OpenAI}}
Gemini{{Gemini}}
Ollama{{Ollama}}
LLMFunc[[ChatEvaluate]]
LLMProm[[LLMPrompt]]
CODBOS[(Persistent<br>Chat objects)]
CODB[(Chat objects)]
PDB[(Prompts)]
CCommand[/Chat command/]
CCommandOutput[/Chat result/]
CIDQ{Chat ID<br>specified?}
CIDEQ{Chat ID<br>exists in DB?}
IngestCODB[Chat objects file<br>ingestion]
UpdateCODB[Chat objects file<br>update]
RECO[Retrieve existing<br>chat object]
COEval[Message<br>evaluation]
PromParse[Prompt<br>DSL spec parsing]
KPFQ{Known<br>prompts<br>found?}
PromExp[Prompt<br>expansion]
CNCO[Create new<br>chat object]
CIDNone["Assume chat ID<br>is 'NONE'"]
subgraph "OS Shell"
CCommand
CCommandOutput
end
subgraph OS file system
CODBOS
end
subgraph PromptProc[Prompt processing]
PDB
LLMProm
PromParse
KPFQ
PromExp
end
subgraph LLMInteract[LLM interaction]
COEval
LLMFunc
Gemini
OpenAI
Ollama
end
subgraph Chatnik backend
IngestCODB
CODB
CIDQ
CIDEQ
CIDNone
RECO
CNCO
UpdateCODB
PromptProc
LLMInteract
end
CCommand --> IngestCODB
CODBOS -.-> IngestCODB
UpdateCODB -.-> CODBOS
IngestCODB -.-> CODB
IngestCODB --> CIDQ
CIDQ --> |yes| CIDEQ
CIDEQ --> |yes| RECO
RECO --> PromParse
COEval --> CCommandOutput
CIDEQ -.- CODB
CIDEQ --> |no| CNCO
LLMFunc -.- CNCO -.- CODB
CNCO --> PromParse --> KPFQ
KPFQ --> |yes| PromExp
KPFQ --> |no| COEval
PromParse -.- LLMProm
PromExp -.- LLMProm
PromExp --> COEval
LLMProm -.- PDB
CIDQ --> |no| CIDNone
CIDNone --> CIDEQ
COEval -.- LLMFunc
COEval --> UpdateCODB
LLMFunc <-.-> OpenAI
LLMFunc <-.-> Gemini
LLMFunc <-.-> Ollama
style PromptProc fill:DimGray,stroke:#333,stroke-width:2px
style LLMInteract fill:DimGray,stroke:#333,stroke-width:2px
Here is the corresponding UML Sequence diagram:
sequenceDiagram
participant CCommand as Chat command
participant IngestCODB as Chat objects file ingestion
participant CODBOS as Chat objects file
participant CODB as Chat objects
participant CIDQ as Chat ID specified?
participant CIDEQ as Chat ID exists in DB?
participant RECO as Retrieve existing chat object
participant PromParse as Prompt DSL spec parsing
participant KPFQ as Known prompts found?
participant PromExp as Prompt expansion
participant COEval as Message evaluation
participant CCommandOutput as Chat result
participant CNCO as Create new chat object
participant CIDNone as Assume chat ID is NONE
participant UpdateCODB as Chat objects file update
participant LLMFunc as LLM Functions
participant LLMProm as LLM Prompts
CCommand->>IngestCODB: Chat command
CODBOS--)IngestCODB: Chat objects file
IngestCODB--)CODB: Chat objects
IngestCODB->>CIDQ: Chat ID specified?
CIDQ-->>CIDEQ: Yes
CIDQ-->>CIDNone: No
CIDNone->>CIDEQ: Assume chat ID is NONE
CIDEQ-->>RECO: Yes
CIDEQ-->>CNCO: No
CIDEQ--)CODB: Chat objects
RECO->>PromParse: Prompt DSL spec parsing
PromParse--)LLMProm: LLM Prompts
CNCO--)LLMFunc: LLM Functions
CNCO--)CODB: Chat objects
CNCO->>PromParse: Prompt DSL spec parsing
PromParse->>KPFQ: Known prompts found?
KPFQ-->>PromExp: Yes
KPFQ-->>COEval: No
PromExp--)LLMProm: LLM Prompts
PromExp->>COEval: Message evaluation
COEval--)LLMFunc: LLM evaluator invocation
LLMFunc--)COEval: Evaluation result
COEval->>UpdateCODB: Chat objects file update
COEval->>CCommandOutput: Chat result
Keeping the persistent chat objects database is a fairly straightforward using the Wolfram Language persistent values system . Efficiency considerations for "using the WL and OS to manage the database" are probably not that important because LLMs invocation is (much) slower in comparison.
[AA1] Anton Antonov, "Chatnik: LLM Host in the Shell --- Part 1: First Examples & Design Principles" , (2026), RakuForPrediction at WordPress .
[AA2] Anton Antonov, "Chatnik: LLM Host in the Shell --- Part 1: First Examples & Design Principles", (2026), PythonForPrediction at WordPress.
[AAp1] Anton Antonov, LLMFunctionObjects, Python package , (2023-2026), GitHub/antononcube . (PyPI.org page.)
[AAp2] Anton Antonov, LLMPrompts, Python package , (2023-2025), GitHub/antononcube . (PyPI.org page.)
[AAp3] Anton Antonov, JupyterChatbook, Python package , (2023-2026), GitHub/antononcube . (PyPI.org page.)
[AAp4] Anton Antonov, Chatnik, Raku package , (2026), GitHub/antononcube .
[CGp1] Connor Gray et al., Chatbook, Wolfram Language paclet, (2023-2024), Wolfram Language Paclet Repository.
[WRIp1] Wolfram Research, Inc., CommandLineParser paclet, (2024), Wolfram Language Paclet Repository.