llm-strings
A lightweight library for message handling with instruction fine-tuned language models.
This module provides two classes: LMString and LMRecordStrings.
-
LMStringis a subclass of Python's built-in string class that overides theformatmethod to provide more informative feedback about missing formatting arguments. -
LMRecordStringsstores multiple string records and formats them according to the input text conversations. You can still format messages after wrapped in a collection, and you have a guarantee over the roles sequence: system (optional), user, assistant, user, assistant,...
Usage
LMString Class
Initialize the LMString
template = LMString("Hello, {name}. Today is {day}.")
# Output: Template String initialized successfully. Required args: {'day', 'name'}
In this example, we create an LMString that expects two arguments: name and day.
Format the LMString
template.format(name="Alice")
# Output: Template String has been partially formatted successfully. Required args now: ['day']
If only a subset of the required arguments are provided, format will return an LMString that only expects the remaining ones.
template.format(name="Alice", day="Monday")
#Output: "Hello, Alice. Today is Monday."
When all required arguments are provided, format will return an ordinary string.
Compose LMString with another string
template @ "How are you?"
# Output: "Hello, {name}. Today is {day}. How are you?"
The @ operator can be used to concatenate an LMString with another string. The resulting LMString will expect all arguments that the original LMString expected.
LMRecordStrings Class
Initialize the LMRecordStrings
record = LMRecordStrings("Hello, {name}.")
In this example, we create an LMRecordStrings that expects one argument: name.
Compose LMRecordStrings with another string or LMRecordStrings
record @ "Today is {day}."
record @ LMRecordStrings("How are you?")
The @ operator can be used to concatenate an LMRecordStrings with another string or LMRecordStrings. The resulting LMRecordStrings will expect all arguments that the original LMRecordStrings expected.
Format the LMRecordStrings
record.format(0, name="Alice")
# Output: 0. Record formatted succesfully. Formatted args: {'name'}. Required args now: set()
The format method formats a specific record in the LMRecordStrings. The first argument is the index of the record to format, and the remaining arguments are the formatting arguments.
record.vformat(name="Alice", day="Monday")
# Output: Records formatted succesfully!
The vformat method formats all records in the LMRecordStrings that expect the provided arguments.
Print the LMRecordStrings
print(record)
The __str__ method returns a string representation of the LMRecordStrings that can be printed to the console. The string representation includes a border and separates each record with a line.
record = LMRecordStrings("Hello, {name}.") @ "Today is {day}." @ LMRecordStrings("How are you?")
record = record.format(0, name="Alice").vformat(day="Monday")
print(record)
This would output:
┌────────────────────────────────────────────────────────────────────────────────────────────────────────────┐
│ 0. System: *** Hello, Alice. *** │
│------------------------------------------------------------------------------------------------------------│
│ 1. User: Today is Monday. │
│------------------------------------------------------------------------------------------------------------│
│ 2. Assistant: How are you? │
└────────────────────────────────────────────────────────────────────────────────────────────────────────────┘
More examples
Optional[System],(User,Asisstant)*n invariant:
# Initializing a conversation with an opening message
conversation = LMRecordStrings() @ "Hello, {name}. How can I assist you today?"
# Adding more records to the conversation
conversation @ "I need assistance with {topic}." @ "Certainly, what specific aspect of {topic} are you interested in?"
# Formatting the first record with the user's name
conversation = conversation.format(0, name="Alice")
# Formatting the second record (user's message) with the topic
conversation = conversation.format(1, topic="Python programming")
# Formatting the third record (assistant's message) with the same topic
conversation = conversation.format(2, topic="Python programming")
# Adding a user reply to the assistant's question
conversation @ "{aspect} of Python programming sounds challenging."
# Formatting the fourth record (user's message) with the specific aspect
conversation = conversation.format(3, aspect="asynchronous code execution")
# Let's add the assistant's final message
conversation @ "I can help you understand {concept} with some examples and explanations."
# Formatting the fifth record (assistant's message) with the specific concept
conversation = conversation.format(4, concept="asynchronous code execution")
# Print the entire conversation
print(conversation)
This will produce an output representing a back-and-forth conversation between the user and the assistant, with roles correctly alternating and messages formatted based on the arguments provided:
┌────────────────────────────────────────────────────────────────────────────────────────────────────────────┐
│ 0. System: *** Hello, Alice. How can I assist you today? *** │
│------------------------------------------------------------------------------------------------------------│
│ 1. User: I need assistance with Python programming. │
│------------------------------------------------------------------------------------------------------------│
│ 2. Assistant: Certainly, what specific aspect of Python programming are you interested in? │
│------------------------------------------------------------------------------------------------------------│
│ 3. User: Asynchronous code execution of Python programming sounds challenging. │
│------------------------------------------------------------------------------------------------------------│
│ 4. Assistant: I can help you understand asynchronous code execution with some examples and explanations. │
└────────────────────────────────────────────────────────────────────────────────────────────────────────────┘
V-Formatting
# Initializing a conversation with an opening message
conversation = LMRecordStrings() @ "Hello, {name}. How can I assist you today?"
# Adding more records to the conversation
conversation @ "I need assistance with {topic}." @ "Certainly, what specific aspect of {topic} are you interested in?"
# Adding a user reply to the assistant's question
conversation @ "{aspect} of {topic} sounds challenging."
# Let's add the assistant's final message
conversation @ "I can help you understand {concept} with some examples and explanations."
# Now, we use vformat to apply the formatting arguments to all records
formatted_conversation = conversation.vformat(
name="Alice",
topic="Python programming",
aspect="asynchronous code execution",
concept="asynchronous code execution"
)
# Print the fully formatted conversation
print(formatted_conversation)
This will produce an output similar to the one before, but this time we are formatting all the records in one go at the end, using the vformat method:
┌────────────────────────────────────────────────────────────────────────────────────────────────────────────┐
│ 0. System: *** Hello, Alice. How can I assist you today? *** │
│------------------------------------------------------------------------------------------------------------│
│ 1. User: I need assistance with Python programming. │
│------------------------------------------------------------------------------------------------------------│
│ 2. Assistant: Certainly, what specific aspect of Python programming are you interested in? │
│------------------------------------------------------------------------------------------------------------│
│ 3. User: Asynchronous code execution of Python programming sounds challenging. │
│------------------------------------------------------------------------------------------------------------│
│ 4. Assistant: I can help you understand asynchronous code execution with some examples and explanations. │
└────────────────────────────────────────────────────────────────────────────────────────────────────────────┘
The vformat method is particularly useful when you have multiple records that require similar arguments, and you want to apply the formatting in a single step. It streamlines the process and makes the code cleaner and more concise, especially when dealing with lengthy conversations.
Metadata
Release files for llm-strings 0.1.3
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | |
|---|---|---|---|---|
| llm_strings-0.1.3.tar.gz | 9.9 kB | Without provenance | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| llm_strings-0.1.3-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 16.5 kB
Release files / llm_strings-0.1.3.tar.gz
| Download URL | llm_strings-0.1.3.tar.gz |
|---|---|
| Size | 9.9 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
7a5cf4f3ab6198b69dc709cd81c4f1833f810714359b90f7a9720783c35956ef
|
|
BLAKE2b-256 checksum How to use checksums |
7cb3faa16e01eb5884e43e30bf15a923fdb7be96fe4b9fbe73cb4cea3a10ec74
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/5.0.0 CPython/3.9.18
|
Release files / llm_strings-0.1.3-py3-none-any.whl
| Download URL | llm_strings-0.1.3-py3-none-any.whl |
|---|---|
| Size | 6.6 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
599a51c6cd8349827bd5d2c0b56de12c7ae3bee3175d42942eb28d257c0a1e50
|
|
BLAKE2b-256 checksum How to use checksums |
83b2b39422ce1780e574bdd5bb2a8469e89fc5cb7e3c17cc13c92030660a70f1
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/5.0.0 CPython/3.9.18
|
Security for llm-strings 0.1.3
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.