New repository structure and refactoring (#17)

Renamed main directory, updated experiments workflow, refactored and cleaned up code

Author: NotBioWaste905

.github/workflows/metrics.yml

@@ -1,11 +1,8 @@
-name: codestyle
+name: tests
 
 on:
   push:
-    branches:
-    - dev
-    - master
-    - test/**
+    branches: '**'
   pull_request:
     branches:
     - dev
@@ -16,7 +13,7 @@ concurrency:
   cancel-in-progress: ${{ github.ref != 'refs/heads/dev' && github.ref != 'refs/heads/master' }}
 
 jobs:
-  lint:
+  test_ubuntu_latest:
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v4
@@ -31,6 +28,6 @@ jobs:
           python -m pip install --upgrade pip poetry
           python -m poetry install --with tests --no-ansi --no-interaction
 
-      - name: run codestyle
+      - name: run tests
         run: |
           python -m poetry run poe test

.github/workflows/tests.yml

@@ -1,6 +1,6 @@
 # Contributing to chatsky-llm-integration
 
-Thank you for your interest in contributing to the dff-llm-integration project! We welcome contributions from the community to help improve and expand this Chatsky LLM-Autoconfig tool.
+Thank you for your interest in contributing to the chatsky-llm-integration project! We welcome contributions from the community to help improve and expand this Chatsky LLM-Autoconfig tool.
 
 ## Getting Started
 
@@ -35,40 +35,18 @@ In order to update versions specified in poetry.lock, run
 poetry update
 ```
 
-## How to test your code
-
-To test your algorithm or pipeline you need to follow these steps:
-
-1. Decorate your class with AlgorithRegistry.register() like so:
-
-```python
-from chatsky_llm_autoconfig.autometrics.registry import AlgorithmRegistry
-
-@AlgorithmRegistry.register(input_type=BaseGraph, output_type=list[Dialogue])
-class DialogueSampler(DialogueGenerator)
-```
-
-2. Make sure that `input_type` and `output_type` are matching with the signature of `.invoke()` method
-3. Run
-
-```bash
-poetry run python dev_packages/chatsky_llm_autoconfig/chatsky_llm_autoconfig/autometrics/run_autometrics.py
-```
-
-4. If metrics didn't drop then everything is fine. Add, commit and push as usual.
-
 ## How to Contribute
 
 1. Make your changes and test hypothesis in the `./experiments` folder as it is described in **Conducting experiments** section
 
-3. Ensure linting using commands as
+2. Ensure linting using commands as
 
-```
-poetry run poe lint
-poetry run poe format
-```
+    ```bash
+    poetry run poe lint
+    poetry run poe format
+    ```
 
-4. Create a pull request with clear description of fixed and features
+3. Create a pull request with clear description of fixes and features
 
 ## Pull Request format
 
@@ -78,8 +56,17 @@ You can always create a draft PR and request review before you request to merge
 ## Conducting experiments
 
 Until any of the code make it way to the main repo it should be tested in `./experiments` folder.
-Each of the experiments must lay in the separate folder with name like `<YYYY.MM.DD>_<experiment_name>`.
-Inside of this directory must be a `report.md` file with results, metrics, future plans and other relevant information.
+Each of the experiments must lay in the separate folder with name like `exp<YYYY>_<MM>_<DD>_<hypothesis>`. Each experiment must be a poetry project that can be done via running either:
+
+```bash
+poetry new --src <experiment_name>
+```
+
+or, alternatively, if you've already created the folder
+
+```bash
+poetry init
+```
 
 **!!! Do not put images into the folder you are commiting, use GoogleDrive instead !!!**
 
@@ -108,10 +95,9 @@ Supported types of graphs:
 
 - [x] chain
 - [x] single cycle
+- [x] multi-cycle graph
+- [x] complex graph with cycles
 
 Currently unsupported types:
 
 - [ ] single node cycle
-- [ ] multi-cycle graph
-- [ ] incomplete graph
-- [ ] complex graph with cycles

CONTRIBUTING.md

@@ -25,10 +25,9 @@ poetry run python <your_file_name>.py
 ## Contents
 
 ```
-./data - Examples, tests and other dialogue data in JSON format
 ./experiments - Test field for experimental features, test data and results
 ./scripts - Here we put scripts needed for `poethepoet` automation (you probably do not need to look inside)
-./dev_packages/chatsky_llm_autoconfig - Directory containing all the code for the `chatsky_llm_autoconfig` module
+./dialogue2graph - Directory containing all the code for the `dialogue2graph` module
 ```
 
 ## Current progress
@@ -43,14 +42,15 @@ Supported types of graphs:
 Currently unsupported types:
 
 - [ ]  single node cycle
-- [ ]  incomplete graph
 
 ## How to use
+
 We provide several of using our library for various tasks.
+
 ### Data generation
 
 ```python
-from chatsky_llm_autoconfig.algorithms.cycle_graph_generation_pipeline import GraphGenerationPipeline
+from dialogue2graph.datasets.complex_dialogues.generation import LoopedGraphGenerator
 from langchain_openai import ChatOpenAI
 
 
@@ -67,10 +67,10 @@ validation_model = ChatOpenAI(
     base_url=os.getenv("OPENAI_BASE_URL"),
     temperature=0
 
-pipeline = GraphGenerationPipeline(
-        generation_model=generation_model,
-        validation_model=validation_model
-    )
+pipeline = LoopedGraphGenerator(
+    generation_model=gen_model,
+    validation_model=val_model,
+)
 
 topics = [
     "technical support conversation",
@@ -83,9 +83,8 @@ topics = [
 successful_generations = []
 
 for topic in topics:
-
     try:
-        result = pipeline(topic)
+        result = pipeline(topic, use_cache=False)
 
         # Check the result type
         if isinstance(result, GraphGenerationResult):
@@ -109,7 +108,7 @@ for topic in topics:
 ### Dialogue sampling
 
 ```python
-from chatsky_llm_autoconfig.algorithms.dialogue_generation import RecursiveDialogueSampler
+from dialogue2graph.pipelines.core.dialogue_sampling import RecursiveDialogueSampler
 from chatsky_llm_autoconfig.graph import Graph
 
 G = Graph(graph_dict={...})
@@ -119,26 +118,6 @@ sampler.invoke(graph=G) #-> list of Dialogue objects
 
 ```
 
-### Graph generation
-### Evaluation
-
-- Generate graph from scratch by topic (input: topic, output: graph) (for dataset generation)
-  - algorithms.topic_graph_generation.CycleGraphGenerator
-
-- change graph without changing graph structure (input: old graph + topic, output: new graph) (for dataset generation)
-
-- sampling from dialogue graph (input: graph, output: dialogue) (for dataset generation)
-  - algorithms.dialogue_generation.DialogueSampler
-
-- augmentation of dialogue (input: dialogue, output: dialogue) (for dataset generation)
-  - algorithms.dialogue_generation.DialogAugmentation
-
-- generate graph from scratch by dialogue (input: dialogue, output: graph) (decisive algorithm)
-  - GeneralGraphGenerator (experiments/2024.11.14_dialogue2graph)
-
-- generate graph based on existing graph (input: old graph + dialog, output: new graph) (extended decision algorithm)
-  - AppendChain (experiments/2024.11.17_concatenating_subchains_prompt)
-
 ## How to contribute?
 
 You can find contribution guideline in [CONTRIBUTING.md](https://github.com/deeppavlov/chatsky-llm-autoconfig/blob/main/CONTRIBUTING.md)