autodocgenerator 0.7.6

This Project helps you to create docs for your projects

Executive Navigation Tree

• 📦 Installation
  • install_workflow'
  • install-scripts
• 🛠️ Manager
  • Using_Manager_Class
  • manager-overview
  • manager‑init
  • manager‑helpers
  • manager‑code‑mix
  • manager‑global‑info
  • manager‑doc‑parts
  • manager‑factory‑doc
  • manager‑cleanup
• 📦 Modules
  • module‑initialisation
  • config_reader
  • run_file
  • base-module
  • custom-module
  • intro-modules
• 📂 Code Mix
  • code‑mix‑overview
  • code‑mix‑api
  • code‑mix‑interaction
• 🗜️ Compressor
  • compressor‑overview
• ⚙️ Core Functions
  • core‑functions
  • doc‑factory
• 🤖 Interaction Model
  • interaction‑model
  • parent-model
  • sync-model
  • async-model
  • gptmodel
  • asyncgptmodel
• 📈 Execution Parts
  • sync-part-doc-generation
  • async-part-doc-generation
• 📊 Orchestration
  • orchestration
• 📜 Logging & Progress
  • logging
  • progress
• 📚 Postprocess
  • postprocess-overview
• 📜 Assumptions & Limits
  • assumptions‑limits
• 📜 History
  • history
• ❗ Exceptions
  • model-exhausted-exception

explain how install workflow with install.ps1 and install.sh scripts for install you should use links irm https://raw.githubusercontent.com/Drag-GameStudio/ADG/main/install.ps1 | iex for powershell and curl -sSL https://raw.githubusercontent.com/Drag-GameStudio/ADG/main/install.sh | bash for linux based systems and also you have to add secret variable to git hub action GROCK_API_KEY with your api key from grock docs https://grockdocs.com to make it work

Parameters required by Manager

• project_path – path to the project (e.g., ".").
• project_settings – an instance of ProjectSettings.
• pcs – an instance of ProjectConfigSettings.
• sync_model – an instance of GPTModel.
• async_model – an instance of AsyncGPTModel.
• ignore_files – list of file names to ignore (list[str]).
• progress_bar – an instance of a progress UI, e.g., ConsoleGtiHubProgress().
• language – language code string, e.g., "en".

Full example of usage (taken from the provided context)

from autodocgenerator.manage import Manager
from autodocgenerator.engine.models.gpt_model import GPTModel, AsyncGPTModel
from autodocgenerator.ui.progress_base import ConsoleGtiHubProgress
from autodocgenerator.preprocessor.settings import ProjectSettings
from autodocgenerator.factory.base_factory import DocFactory
from .config_reader import Config, read_config, ProjectConfigSettings
from autodocgenerator.engine.config.config import API_KEY

# Load configuration
with open("autodocconfig.yml", "r", encoding="utf-8") as file:
    config_data = file.read()
config: Config = read_config(config_data)

# Extract needed objects from config
project_settings = config.get_project_settings()
doc_factory, intro_factory = config.get_doc_factory()

# Prepare GPT models
sync_model = GPTModel(API_KEY, use_random=False)
async_model = AsyncGPTModel(API_KEY)

# Other arguments
project_path = "."                     # current directory
pcs: ProjectConfigSettings = config.pcs
ignore_list = config.ignore_files
progress = ConsoleGtiHubProgress()
language = "en"

# Create the Manager instance
manager = Manager(
    project_path,
    project_settings,
    pcs,
    sync_model=sync_model,
    async_model=async_model,
    ignore_files=ignore_list,
    progress_bar=progress,
    language=language,
)

# Example operations using the manager
manager.generate_code_file()
manager.generate_global_info_file(use_async=False, max_symbols=8000)
manager.generete_doc_parts(use_async=False, max_symbols=5000)
manager.factory_generate_doc(doc_factory)
manager.factory_generate_doc(intro_factory)
manager.clear_cache()
output_doc = manager.read_file_by_file_key("output_doc")

The autodocconfig.yml file is a YAML document that defines the configuration for the Auto Doc Generator.
Based on the provided context, the following top‑level keys are recognized:

Key	Type	Description	Example
project_name	string	Name of the project that will appear in the generated documentation.	project_name: "Auto Doc Generator"
language	string	Language code for the documentation (default en).	language: "en"
project_settings	mapping	Settings that affect the documentation generation process.	project_settings: save_logs: true log_level: 2
project_additional_info	mapping	Arbitrary key‑value pairs that are added to the project information section.	project_additional_info: global idea: "This project was created to help developers make documentations for them projects"
ignore_files	list of strings (optional)	Glob patterns for files/folders that should be ignored during scanning. If omitted the default list from Config is used.	ignore_files: - "*.md" - "tests/**"
custom_descriptions	list of strings	Custom textual descriptions that will be turned into modules by the generator. Each entry is processed as a CustomModule.	custom_descriptions: - "explain how install workflow…" - "how to use Manager class…" - "explain how to write autodocconfig.yml file…"

Typical structure

project_name: "My Project"
language: "en"

project_settings:
  save_logs: true
  log_level: 2

project_additional_info:
  description: "Short description of the project"
  author: "Your Name"

ignore_files:
  - "*.md"
  - "docs/**"

custom_descriptions:
  - "First custom description"
  - "Second custom description"

📦 autodocgenerator package initialisation

The autodocgenerator/__init__.py file is executed on every import of the ADG library.
Its sole responsibilities are:

1. Signal successful import – the print("ADG") statement writes a short banner to stdout (useful during debugging or when the package is invoked as a script).
2. Expose a ready‑to‑use logger – it imports the concrete logging classes from autodocgenerator.ui.logging and creates a module‑level logger instance:

   logger = BaseLogger()
   logger.set_logger(BaseLoggerTemplate())
   

   This singleton is shared across the entire code‑base, so any sub‑module can simply from autodocgenerator import logger and emit InfoLog, WarningLog or ErrorLog messages without configuring a logger themselves.

Interaction with the rest of the system

• UI → Engine → Factory components all import logger to report progress, configuration problems, or fatal errors.
• The logger respects the global settings defined in autodocconfig.yml (e.g., save_logs and log_level) because BaseLoggerTemplate reads those values during its own initialisation.

Assumptions & side‑effects

• The environment must have the autodocgenerator.ui.logging module available; otherwise an ImportError aborts package loading.
• The print side‑effect writes to standard output every time the package is imported – in production pipelines this can be silenced by redirecting stdout or by removing the line.

Overall, this file provides a tiny but essential bootstrap: a consistent logger and a visual cue that the Auto Doc Generator library has been loaded.

autodocgenerator.auto_runner.config_reader – Configuration Bootstrap

This module translates autodocconfig.yml into runtime objects used throughout the generator.

• ProjectConfigSettings – holds global logger flags (save_logs, log_level). load_settings() copies any YAML keys into the instance via setattr.

• Config – central holder for:

  • ignore_files – default file‑patterns excluded from scanning.
  • language, project_name, project_additional_info.
  • custom_modules – list of CustomModule objects supplied by the user.
  • pcs – an attached ProjectConfigSettings.
    It supplies fluent setters (set_language, add_ignore_file, …) and two factories:
  • get_project_settings() → builds a ProjectSettings object populated with extra info.
  • get_doc_factory() → creates the main DocFactory (with custom modules) and an intro factory (IntroLinks, optionally IntroText).

• read_config(file_data: str) -> Config – parses the YAML, fills a Config instance, and returns it.

  • Reads ignore_files, language, project_name, project_additional_info, project_settings, and custom_descriptions.
  • Instantiates CustomModule for each user‑provided description.

Interaction – All higher‑level components (run_file, Manager, factories) import Config to obtain:

• the ignore list passed to Manager,
• the ProjectSettings for metadata,
• the prepared DocFactory objects, and
• the logger configuration (pcs).

Assumptions / Side‑effects

• The YAML file is syntactically correct; otherwise yaml.safe_load raises.
• CustomModule accepts a raw description dict.
• No I/O is performed here – the caller must read the file (as run_file does).

---

autodocgenerator.auto_runner.run_file – Execution Entrypoint

run_file.py glues the configuration, AI models, and generation pipeline together.

• gen_doc(...) -> str

  • Instantiates GPTModel (sync) and AsyncGPTModel (async) using the global API_KEY.
  • Builds a Manager with project path, ProjectSettings, ProjectConfigSettings, the models, ignore list, a ConsoleGtiHubProgress bar, and language.
  • Calls the manager’s stages in order:
    1. generate_code_file – extracts source code snippets.
    2. generate_global_info_file – creates a high‑level overview (max 8 000 symbols).
    3. generete_doc_parts – produces modular doc chunks (max 5 000 symbols).
    4. factory_generate_doc – runs both the main and intro DocFactory objects.
    5. clear_cache – removes temporary artefacts.
  • Returns the assembled documentation via manager.read_file_by_file_key("output_doc").

• __main__ block – reads autodocconfig.yml, builds a Config via read_config, extracts ProjectSettings and factories, then calls gen_doc and stores the result in output_doc.

Interaction –

• Relies on Config (above) for all runtime parameters.
• Uses the singleton logger from autodocgenerator.ui.logging (imported but not explicitly called here; internal Manager and factories emit logs).
• Progress is visualised through ConsoleGtiHubProgress, a subclass of BaseProgress.

Assumptions –

• API_KEY is defined in engine.config.config.
• Required packages (rich, yaml, GPT model wrappers) are installed.
• The current working directory is the project root (passed as ".").

Together, these two files form the bootstrap that reads user configuration, prepares AI back‑ends, and drives the full documentation generation pipeline.

autodocgenerator.engine.config.config – Prompt Templates & Global Settings

Responsibility
This module centralises all static prompt texts and runtime constants used by the AutoDoc generation pipeline. The strings (BASE_SYSTEM_TEXT, BASE_PART_COMPLITE_TEXT, BASE_INTRODACTION_CREATE_TEXT, BASE_INTRO_CREATE, BASE_SETTINGS_PROMPT) are injected into LLM calls to steer behaviour (e.g., step‑by‑step code analysis, documentation scaffolding, navigation‑tree creation).

Key Elements

Symbol	Type	Purpose
BASE_SYSTEM_TEXT	str	System‑level instruction for incremental code‑snippet analysis.
BASE_PART_COMPLITE_TEXT	str	Template for the “write documentation” stage.
BASE_INTRODACTION_CREATE_TEXT	str	Prompt that generates an executive navigation tree from Markdown links.
BASE_INTRO_CREATE	str	Prompt for a full project overview (title, goal, core logic, etc.).
BASE_SETTINGS_PROMPT	str	Prompt that establishes a persistent project knowledge base for later interactions.
get_BASE_COMPRESS_TEXT(start, power)	Callable[[int, int], str]	Factory that builds a compression‑task prompt, scaling the allowed summary length (~start/power chars) and embedding a strict usage‑example skeleton.
API_KEY	str	Obtained from environment (.env) – required by the GPT model wrappers.
MODELS_NAME	list[str]	Default model identifiers the system can select from.

Interaction with the System

• LLM factories / managers import these constants to build prompts for GPTModel, AsyncGPTModel, and the various DocFactory objects.
• read_config (in the configuration loader) does not use them directly, but the generated Config object later supplies them to the pipeline via Manager.
• run_file pulls API_KEY and MODELS_NAME to instantiate model wrappers before constructing the Manager.

Assumptions & Side‑effects

• The .env file is present and defines API_KEY; otherwise an exception aborts start‑up.
• Prompt strings are treated as raw literals – no runtime formatting occurs here.
• get_BASE_COMPRESS_TEXT only returns a formatted string; it does not perform any I/O.

Typical Usage Example

from autodocgenerator.engine.config.config import (
    BASE_SYSTEM_TEXT,
    get_BASE_COMPRESS_TEXT,
    API_KEY,
    MODELS_NAME,
)

# Build a compression prompt for a 5000‑char snippet, aiming for ~500‑char summary
compress_prompt = get_BASE_COMPRESS_TEXT(start=5000, power=10)

# Pass prompts to the LLM wrapper (pseudo‑code)
gpt = GPTModel(api_key=API_KEY, model_name=MODELS_NAME[0])
response = gpt.complete(system=BASE_SYSTEM_TEXT, user=compress_prompt)

ModelExhaustedException
ModelExhaustedException is a custom sentinel raised when the runtime‑selected list of model identifiers (regen_models_name) becomes empty. It propagates up to the calling factory or UI layer, signalling that no fallback LLM is available and that request processing must abort.

---

History
A lightweight container for the conversation payload sent to Groq.

• Constructor History(system_prompt: str = BASE_SYSTEM_TEXT) – starts with an empty list and optionally injects the system prompt.
• add_to_history(role, content) – appends a dict {"role": role, "content": content}.
• Interaction: Instances are shared with Model/AsyncModel so every generated answer can be stored and later reused when with_history=True.

---

ParentModel
Base class that prepares common state for concrete model wrappers.

• Stores the API key, a History object, and an index (current_model_index).
• Copies the global MODELS_NAME list, shuffles it when use_random=True, and exposes the ordered list as regen_models_name.
• Assumption: MODELS_NAME contains at least one valid Groq model identifier; otherwise ModelExhaustedException will be raised later.

---

Model (synchronous)
Derives from ParentModel and implements the high‑level public API used by the rest of the system.

• generate_answer(with_history=True, prompt=None) → str – abstract placeholder overridden by GPTModel.
• get_answer_without_history(prompt) – convenience wrapper that forwards a raw message list directly to generate_answer.
• get_answer(prompt) – records the user message, calls generate_answer (which uses history by default), stores the assistant reply, and returns it.
• Side‑effects: Mutates self.history by appending user and assistant entries.

---

AsyncModel (asynchronous)
Mirrors Model but with async signatures, enabling non‑blocking UI or batch pipelines.

• Same three public helpers (generate_answer, get_answer_without_history, get_answer) but all awaitable.
• History handling is identical to the sync version.

---

GPTModel – concrete synchronous wrapper around Groq’s SDK.

• Instantiates self.client = Groq(api_key=self.api_key) and a BaseLogger.
• generate_answer
  1. Logs start.
  2. Picks messages from self.history or the supplied prompt.
  3. Loops over regen_models_name trying each model until a successful chat.completions.create call returns.
  4. On failure logs a warning, advances current_model_index, and retries; if the list empties, raises ModelExhaustedException.
  5. Extracts result = chat_completion.choices[0].message.content, logs it, and returns the string.
• Inputs: with_history flag, optional raw prompt.
• Outputs: generated answer string.
• Side‑effects: logging, possible index mutation, no direct history mutation (handled by Model.get_answer).

---

AsyncGPTModel – asynchronous counterpart of GPTModel.

• Uses AsyncGroq client and the same logging strategy.
• The generate_answer coroutine follows the identical retry logic, awaiting self.client.chat.completions.create.
• Returns the answer string once the awaitable completes.

---

Interaction Overview

1. Factory / UI creates a GPTModel or AsyncGPTModel, passing the global API_KEY and optionally a custom History.
2. Callers invoke get_answer / get_answer_without_history.
3. The wrapper builds the message payload, selects a model (randomised order), contacts Groq, logs each step, and returns the assistant response.
4. If every candidate model fails, ModelExhaustedException bubbles up for graceful error handling.

BaseModule (abstract)
Defines the contract for all documentation‑generation plug‑ins.

• Responsibility – Provide a generate(info, model) method that receives a parsed source‑code dictionary (info) and a concrete Model (sync or async) to query LLMs.
• Interaction – Sub‑classes are instantiated by DocFactory and called sequentially. No side effects beyond returning a string fragment.

DocFactory
Orchestrates module execution and aggregates results.

• Constructor – Accepts any number of BaseModule instances; stores them in self.modules.
• generate_doc(info, model, progress) –
  1. Creates a sub‑task in the supplied BaseProgress UI component.
  2. Iterates over self.modules, invoking module.generate(info, model).
  3. Concatenates each fragment (module_result) with double new‑lines.
  4. Logs success and raw output via BaseLogger.
  5. Updates progress after each module and finally removes the sub‑task.
• Outputs – The full assembled documentation string.
• Side effects – UI progress updates, structured logging, no mutation of info or model.

CustomModule (general‑purpose)

• Purpose – Produce a user‑defined description block.
• Inputs – info["code_mix"] (mixed source code), info["language"], and the constructor‑provided discription text.
• Logic – Splits the code into ≤ 7 000‑symbol chunks (split_data), then calls generete_custom_discription with the chunks, the LLM model, the custom description, and the target language. Returns the resulting markdown/HTML fragment.

IntroLinks & IntroText

• IntroLinks – Extracts every HTML anchor from info["full_data"] (get_all_html_links), prints the list (debug), and asks the LLM (get_links_intro) to compose a brief “quick‑links” section in the desired language.
• IntroText – Requests a high‑level introduction for the project (get_introdaction) based on info["global_data"] and the language.
  Both modules return plain strings that are later stitched by DocFactory.

Manager – Orchestrator of the documentation pipeline
The Manager class is the central coordinator that ties together all preprocessing, LLM‑driven generation, post‑processing and UI‑feedback components of AutoDocGenerator. It receives a project path and configuration objects, prepares a cache directory, and exposes high‑level actions that are called by the CLI / UI.

---

Construction & state

def __init__(self,
             project_directory: str,
             project_settings: ProjectSettings,
             pcs: ProjectConfigSettings,
             sync_model: Model = None,
             async_model: AsyncModel = None,
             ignore_files: list = [],
             language: str = "en",
             progress_bar: BaseProgress = BaseProgress())

• Inputs – paths, settings, optional synchronous/asynchronous LLM wrappers, file‑ignore list, UI progress object.
• Side‑effects – creates <project>/ .auto_doc_cache if missing and configures a file logger (FileLoggerTemplate) that writes to report.log.
• Assumptions – at least one of sync_model/async_model is supplied; pcs.log_level is a valid logging level.

---

Helper utilities

• read_file_by_file_key(file_key) – opens a cached file (code mix, global info, output, logs) and returns its text.
• get_file_path(file_key) – builds an absolute path inside the cache folder using FILE_NAMES.

Both helpers are used by every public method, ensuring a single source of truth for cache locations.

---

generate_code_file – building the raw source snapshot

1. Logs start/end via BaseLogger.
2. Instantiates CodeMix(project_directory, ignore_files).
3. Calls CodeMix.build_repo_content(target_path) which walks the repository, concatenates source files (respecting ignore_files) and writes the result to code_mix.txt.
4. Updates the UI progress bar.

Outputs: a plain‑text “code mix” file used by later stages.

---

generate_global_info_file – placeholder for project‑wide summary

Current implementation reads the previously generated code mix, then writes a stub ("ss") to global_info.md.
Intended flow (commented out) would:

1. Split the code mix (split_data) into ≤ max_symbols chunks.
2. Choose sync_model or async_model.
3. Call compress_to_one to let the LLM synthesize a high‑level description of the whole project.

The method also advances the progress bar.

---

generete_doc_parts – chunk‑wise documentation creation

• Reads global_info and the full code mix.
• Depending on use_async, runs either async_gen_doc_parts (via asyncio.run) or gen_doc_parts. Both functions:
  • Split the source into ≤ max_symbols pieces.
  • Prompt the LLM (sync or async) to produce module‑level docs.
  • Return a concatenated markdown string.
• Writes the result to output_doc.md and updates the progress UI.

Outputs: a first‑pass documentation file containing generated sections for every code chunk.

---

factory_generate_doc – post‑processing with a DocFactory

1. Loads global_info, the current output_doc, and the original code_mix.
2. Packages them into an info dict (language, global_data, full_data, code_mix).
3. Logs a detailed start message, listing the concrete module classes contained in the supplied DocFactory.
4. Calls doc_factory.generate_doc(info, self.sync_model, self.progress_bar).
   • The factory runs a configurable pipeline of modules (e.g., IntroLinks, IntroText, CustomModule) that may add introductions, quick‑link sections, or custom markdown.
5. Prepends the factory output to the existing documentation and rewrites output_doc.md.
6. Updates the progress bar.

Side‑effects: modifies the final doc file and produces additional log entries.

---

clear_cache – optional log cleanup

If pcs.save_logs is False, removes report.log from the cache folder. No other files are touched.

---

Interaction diagram (textual)

[CLI/UI] → Manager
    │
    ├─> CodeMix (preprocessor) → code_mix.txt
    ├─> split_data / compress_to_one (preprocessor) → global_info.md
    ├─> gen_doc_parts / async_gen_doc_parts (preprocessor) → output_doc.md
    ├─> DocFactory (factory) → IntroLinks, IntroText, CustomModule → enrich output_doc.md
    └─> BaseLogger / BaseProgress → console & file logs

All heavy LLM calls are delegated to the Model/AsyncModel objects; the manager merely forwards them and handles I/O, logging and progress reporting.

---

This documentation covers the Manager component only; other modules (spliter, compressor, postprocess, UI) are described elsewhere.

CodeMix – repository‑wide source collector

Responsibility – Walk a project directory, filter out unwanted files/folders and emit a single code‑mix text file that contains a tree view of the repository followed by the raw contents of every kept file. This file is the primary input for later preprocessing (splitting, compression) and documentation generation stages.

Public API

Member	Type / signature	Description
__init__(self, root_dir=".", ignore_patterns=None)	`root_dir: str	Path– base folder;ignore_patterns: list[str]` – glob patterns.
should_ignore(self, path: Path) -> bool	Returns True if path (relative to root_dir) matches any pattern (full path, basename, or any path component).
build_repo_content(self, output_file="repomix-output.txt")	Writes the mixed file.	1️⃣ Emits “Repository Structure” with indented tree. 2️⃣ Inserts a separator line. 3️⃣ For each non‑ignored file writes <file path="…"> header and its UTF‑8 text (errors ignored). Logs ignored items at level 1.

Interaction with the system

• Inputs – Physical files under root_dir; optional ignore_patterns.
• Outputs – output_file (plain‑text “code‑mix”); log entries via BaseLogger.
• Consumers – Manager.generate_global_info_file, Manager.generete_doc_parts, and any downstream component that expects a single mixed source file.
• Side effects – File creation on disk, logging; no mutation of source files.

Assumptions & limits

• All source files are readable as UTF‑8 (binary files are filtered out by patterns).
• Ignoring is purely pattern‑based; complex VCS‑aware rules must be expressed as globs.
• The separator ("="*20) marks the boundary for downstream splitters.

---

The ignore_list defined at module level provides sensible defaults (virtual envs, caches, compiled artefacts, markdown docs, etc.).

Compressor module – high‑level view

The compressor prepares source‑code chunks for downstream processing (e.g., global‑info generation). It reduces large texts by sending them to a LLM ( Model / AsyncModel ) with a system prompt derived from ProjectSettings and a configurable compress_power. The output is a shortened representation that still preserves structural clues.

---

Key functions

Function	Responsibility	Main I/O	Side‑effects
compress	Build a three‑message prompt and obtain a single LLM answer.	data: str → str	None (network call only).
compress_and_compare	Synchronously batch‑compress compress_power files, concatenate results, and report progress.	list[str] → list[str] (≈ len/ compress_power items)	Updates BaseProgress.
async_compress	Same as compress but guarded by an asyncio.Semaphore and updates progress asynchronously.	str → str (awaitable)	None besides progress update.
async_compress_and_compare	Parallel version of compress_and_compare using a semaphore (default 4 workers).	list[str] → list[str]	Progress sub‑task handling.
compress_to_one	Repeatedly compresses until a single aggregated string remains; switches to async mode if requested.	list[str] → str	May invoke asyncio.run.
generate_describtions_for_code	Sends each compressed block to the LLM with a strict “describe API” prompt, collects markdown‑ready descriptions.	list[str] → list[str]	Progress updates; no file I/O.

---

Interaction with the rest of the system

• Inputs – Raw source‑code strings produced by the collector stage.
• Outputs – Compressed text blobs consumed by Manager.generate_global_info_file and later splitters; description strings fed to the doc‑generator stage.
• Dependencies –
  • ProjectSettings.prompt (system context).
  • get_BASE_COMPRESS_TEXT (template for compression intensity).
  • BaseProgress for UI feedback.
  • LLM model interfaces (Model, AsyncModel).

The module is pure‑logic: it never writes files, only returns strings and updates the progress UI.

---

Assumptions & limits

• All incoming texts are UTF‑8 readable; binary artefacts must be filtered earlier.
• Compression quality is driven solely by compress_power (default 4) and the static prompt template – no VCS‑aware heuristics.
• Asynchronous workers are capped at four concurrent calls to avoid overwhelming the LLM endpoint.

---

Use this module when you need to shrink large codebases into a single “code‑mix” representation before documentation generation.

postprocess – Final‑stage documentation polishing

The postprocess module is the last step of the autodocgenerator pipeline.
After the raw “code‑mix” markdown has been created it extracts headings and HTML anchors, asks the LLM to craft a short introduction that lists the available sections, and optionally generates a custom description for a user‑supplied topic.

Responsibility

• Parse markdown headers (##) → generate Markdown‑compatible anchors.
• Scan existing <a name=…> tags → collect internal HTML links.
• Build LLM prompts (using BASE_INTRODACTION_CREATE_TEXT / BASE_INTRO_CREATE) and retrieve generated introductory text.
• Provide a helper (generete_custom_discription) that iteratively asks the model for a concise description until a non‑empty answer is obtained.

Interaction with the system

• Relies on the generic Model interface (model.get_answer_without_history).
• Uses configuration constants from engine.config.config.
• Emits progress information through ui.logging.BaseLogger (InfoLog).

Key functions

Function	Purpose	Input / Output
generate_markdown_anchor(header:str) -> str	Normalises a header to a GitHub‑style #anchor.	Header text → #slug
get_all_topics(data:str) -> tuple[list[str], list[str]]	Finds every ## heading, returns titles and their markdown anchors.	Full markdown → (titles, anchors)
get_all_html_links(data:str) -> list[str]	Extracts internal <a name=…> links (max 25 chars) and logs the count.	Full markdown → list of #name
get_links_intro(links:list[str], model:Model, language:str="en") -> str	Calls the LLM to produce an introductory paragraph that enumerates the supplied links.	Links + model → generated text
get_introdaction(global_data:str, model:Model, language:str="en") -> str	Generates a generic introduction for the whole documentation block.	Full doc + model → intro
generete_custom_discription(splited_data:str, model:Model, custom_description:str, language:str="en") -> str	Loops over split fragments, asking the model for a strict description of a custom topic; stops on the first non‑empty answer.	Fragments + description + model → description or empty string

Assumptions & side effects

• Input markdown uses \n## for top‑level sections and <a name=…> for anchors.
• The LLM is expected to respect the Strict Rules prompt; otherwise the function may return "!noinfo" or "No information found".
• Logging is performed synchronously; no external state is mutated besides the logger.

This module therefore “tidies up” the generated documentation, making it navigable and providing human‑readable introductions before the final output is written to disk.

Data Splitting – split_data

Responsibility – Breaks a large source‑code string into chunks that respect max_symbols (≈ token limit of the LLM).
Interaction – Called by the orchestration functions (gen_doc_parts, async_gen_doc_parts) to produce the list splited_data.
Logic – Iteratively trims any chunk > 1.5 × max_symbols by inserting a half‑size slice, then packs the resulting pieces into split_objects while ensuring no object exceeds 1.25 × max_symbols.
Assumptions – Input is a plain‑text code dump; max_symbols is a positive integer.
Outputs – List[str] where each element is safe to send to the LLM.

Synchronous Part Documentation – write_docs_by_parts

Responsibility – Sends a single code chunk to the LLM and returns the generated markdown.
Interaction – Consumes Model.get_answer_without_history; receives the previous part’s output (prev_info) to maintain context.
Key Steps –

1. Build a system‑prompt that forces the requested language and injects BASE_PART_COMPLITE_TEXT.
2. Optionally append the previous documentation fragment.
3. Append the user‑prompt containing the code part (twice, per original design).
4. Strip surrounding triple‑backticks from the LLM answer.
   Inputs – part: str, model: Model, global_info: str (currently unused), optional prev_info, language.
   Outputs – Cleaned documentation string for the part.

Asynchronous Part Documentation – async_write_docs_by_parts

Responsibility – Same as the sync version but runs concurrently under a semaphore.
Interaction – Uses AsyncModel.get_answer_without_history; calls update_progress after each LLM response.
Side Effects – Logs progress and may raise if the semaphore limit is exceeded.

Orchestration – gen_doc_parts / async_gen_doc_parts

Responsibility – Drives full‑document generation by:

1. Splitting the whole code base (split_data).
2. Serially (or concurrently) invoking the part‑generation helpers.
3. Accumulating results, keeping a 3 000‑character tail (result = result[-3000:]) to provide context for the next chunk.
4. Updating a BaseProgress sub‑task for UI feedback.
   Interactions – Relies on the logger (BaseLogger), progress bar, and the appropriate model interface.
   Outputs – The complete documentation string for the supplied code base.

These components together enable the AutoDocGenerator to respect LLM token limits while producing coherent, language‑specific documentation in both synchronous and asynchronous pipelines.

AutoDocGenerator – UI & Setup Documentation

Logging subsystem (ui/logging.py)

• Responsibility – Provides a lightweight, extensible logging façade used throughout the generator (e.g., during parsing, model calls, file writes).
• Core classes
  • BaseLog: stores message and numeric level; _log_prefix adds a timestamp.
  • ErrorLog, WarningLog, InfoLog: override format() to prepend severity tags.
  • BaseLoggerTemplate: holds a log_level filter and routes logs to print().
  • FileLoggerTemplate: same API but appends formatted messages to a user‑specified file.
  • BaseLogger (singleton): holds the active logger_template; set_logger() swaps implementations, log() forwards to global_log() which respects the filter.
• Interaction – UI components (e.g., progress_base, CLI) call BaseLogger().log(ErrorLog(...)); the concrete template decides where the text ends up.

Progress subsystem (ui/progress_base.py)

• Responsibility – Visual feedback for multi‑step documentation generation, abstracted to allow Rich‑based or plain‑console rendering.
• Key abstractions
  • BaseProgress: defines the interface (create_new_subtask, update_task, remove_subtask).
  • LibProgress: wraps a rich.progress.Progress object, creates a “General progress” task and optional sub‑tasks, advancing the appropriate bar on each update_task().
  • ConsoleTask: tiny helper that prints start and percentage updates.
  • ConsoleGtiHubProgress: implements BaseProgress using ConsoleTask; useful when Rich is unavailable (e.g., CI logs).
• Interaction – The main generation pipeline instantiates either LibProgress or ConsoleGtiHubProgress and calls the three interface methods as stages (parsing, model inference, file writing) progress.

Project bootstrap scripts

• install.ps1 (PowerShell) and install.sh (Bash) create a GitHub‑Actions workflow (.github/workflows/autodoc.yml) and a minimal autodocconfig.yml.
• They ensure the workflow directory exists, write static YAML content with proper escaping, inject the current folder name as project_name, and confirm success via console output.
• These scripts are the entry point for users to enable the AutoDocGenerator CI integration with a single command.

---

All components cooperate: the logger records events, the progress visualizer reflects the pipeline state, and the install scripts set up CI to run the generator automatically.

It looks like you’ve pasted the pyproject.toml file twice, but the request was to write documentation for a specific code fragment (the “last part” of the documentation that sets up the CI‑integration entry point).

Could you share the actual Python script / module you’d like documented (e.g., the install‑or‑setup script that prints the folder name as project_name and confirms success on the console)? Once I have that snippet I can produce the concise, linked documentation you need.