janus-llm 4.1.0__tar.gz → 4.3.1__tar.gz

{janus_llm-4.1.0 → janus_llm-4.3.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: janus-llm
-Version: 4.1.0
+Version: 4.3.1
 Summary: A transcoding library using LLMs.
 Home-page: https://github.com/janus-llm/janus-llm
 License: Apache 2.0
@@ -23,20 +23,28 @@ Requires-Dist: langchain-anthropic (>=0.1.15,<0.2.0)
 Requires-Dist: langchain-community (>=0.2.0,<0.3.0)
 Requires-Dist: langchain-core (>=0.2.0,<0.3.0)
 Requires-Dist: langchain-openai (>=0.1.8,<0.2.0)
+Requires-Dist: langchain-unstructured (>=0.1.2,<0.2.0)
 Requires-Dist: nltk (>=3.8.1,<4.0.0)
 Requires-Dist: numpy (>=1.24.3,<2.0.0)
 Requires-Dist: openai (>=1.14.0,<2.0.0)
+Requires-Dist: pi-heif (>=0.20.0,<0.21.0)
 Requires-Dist: py-readability-metrics (>=1.4.5,<2.0.0)
 Requires-Dist: py-rouge (>=1.1,<2.0)
+Requires-Dist: pytesseract (>=0.3.13,<0.4.0)
 Requires-Dist: python-dotenv (>=1.0.0,<2.0.0)
 Requires-Dist: rich (>=13.7.1,<14.0.0)
 Requires-Dist: sacrebleu (>=2.4.1,<3.0.0)
+Requires-Dist: scikit-learn (>=1.5.2,<2.0.0)
 Requires-Dist: sentence-transformers (>=2.6.1,<3.0.0) ; extra == "hf-local" or extra == "all"
+Requires-Dist: tesseract (>=0.1.3,<0.2.0)
 Requires-Dist: text-generation (>=0.6.0,<0.7.0)
 Requires-Dist: tiktoken (>=0.7.0,<0.8.0)
 Requires-Dist: transformers (>=4.31.0,<5.0.0)
 Requires-Dist: tree-sitter (>=0.21.0,<0.22.0)
 Requires-Dist: typer (>=0.9.0,<0.10.0)
+Requires-Dist: unstructured (>=0.15.9,<0.16.0)
+Requires-Dist: unstructured-inference (>=0.7.36,<0.8.0)
+Requires-Dist: unstructured-pytesseract (>=0.3.13,<0.4.0)
 Project-URL: Documentation, https://janus-llm.github.io/janus-llm
 Project-URL: Repository, https://github.com/janus-llm/janus-llm
 Description-Content-Type: text/markdown

{janus_llm-4.1.0 → janus_llm-4.3.1}/janus/__init__.py

@@ -5,7 +5,7 @@ from langchain_core._api.deprecation import LangChainDeprecationWarning
 from janus.converter.translate import Translator
 from janus.metrics import *  # noqa: F403
 
-__version__ = "4.1.0"
+__version__ = "4.3.1"
 
 # Ignoring a deprecation warning from langchain_core that I can't seem to hunt down
 warnings.filterwarnings("ignore", category=LangChainDeprecationWarning)

{janus_llm-4.1.0 → janus_llm-4.3.1}/janus/cli.py

@@ -13,10 +13,14 @@ from rich.console import Console
 from rich.prompt import Confirm
 from typing_extensions import Annotated
 
+import janus.refiners.refiner
+import janus.refiners.uml
 from janus.converter.aggregator import Aggregator
 from janus.converter.converter import Converter
 from janus.converter.diagram import DiagramGenerator
 from janus.converter.document import Documenter, MadLibsDocumenter, MultiDocumenter
+from janus.converter.evaluate import InlineCommentEvaluator, RequirementEvaluator
+from janus.converter.partition import Partitioner
 from janus.converter.requirements import RequirementsDocumenter
 from janus.converter.translate import Translator
 from janus.embedding.collections import Collections
@@ -44,7 +48,6 @@ from janus.llm.models_info import (
     openai_models,
 )
 from janus.metrics.cli import evaluate
-from janus.refiners.refiner import REFINERS
 from janus.utils.enums import LANGUAGES
 from janus.utils.logger import create_logger
 
@@ -69,6 +72,18 @@ with open(db_file, "r") as f:
 collections_config_file = Path(db_loc) / "collections.json"
 
 
+def get_subclasses(cls):
+    return set(cls.__subclasses__()).union(
+        set(s for c in cls.__subclasses__() for s in get_subclasses(c))
+    )
+
+
+REFINER_TYPES = get_subclasses(janus.refiners.refiner.JanusRefiner).union(
+    {janus.refiners.refiner.JanusRefiner}
+)
+REFINERS = {r.__name__: r for r in REFINER_TYPES}
+
+
 def get_collections_config():
     if collections_config_file.exists():
         with open(collections_config_file, "r") as f:
@@ -113,7 +128,7 @@ embedding = typer.Typer(
 
 def version_callback(value: bool) -> None:
     if value:
-        from janus import __version__ as version
+        from . import __version__ as version
 
         print(f"Janus CLI [blue]v{version}[/blue]")
         raise typer.Exit()
@@ -244,22 +259,23 @@ def translate(
             click_type=click.Choice(list(CUSTOM_SPLITTERS.keys())),
         ),
     ] = "file",
-    refiner_type: Annotated[
-        str,
+    refiner_types: Annotated[
+        list[str],
         typer.Option(
             "-r",
             "--refiner",
-            help="Name of custom refiner to use",
+            help="List of refiner types to use. Add -r for each refiner to use in\
+                refinement chain",
             click_type=click.Choice(list(REFINERS.keys())),
         ),
-    ] = "none",
+    ] = ["JanusRefiner"],
     retriever_type: Annotated[
         str,
         typer.Option(
             "-R",
             "--retriever",
             help="Name of custom retriever to use",
-            click_type=click.Choice(["active_usings"]),
+            click_type=click.Choice(["active_usings", "language_docs"]),
         ),
     ] = None,
     max_tokens: Annotated[
@@ -272,6 +288,7 @@ def translate(
         ),
     ] = None,
 ):
+    refiner_types = [REFINERS[r] for r in refiner_types]
     try:
         target_language, target_version = target_lang.split("-")
     except ValueError:
@@ -296,7 +313,7 @@ def translate(
         db_path=db_loc,
         db_config=collections_config,
         splitter_type=splitter_type,
-        refiner_type=refiner_type,
+        refiner_types=refiner_types,
         retriever_type=retriever_type,
     )
     translator.translate(input_dir, output_dir, overwrite, collection)
@@ -402,22 +419,23 @@ def document(
             click_type=click.Choice(list(CUSTOM_SPLITTERS.keys())),
         ),
     ] = "file",
-    refiner_type: Annotated[
-        str,
+    refiner_types: Annotated[
+        list[str],
         typer.Option(
             "-r",
             "--refiner",
-            help="Name of custom refiner to use",
+            help="List of refiner types to use. Add -r for each refiner to use in\
+                refinement chain",
             click_type=click.Choice(list(REFINERS.keys())),
         ),
-    ] = "none",
+    ] = ["JanusRefiner"],
     retriever_type: Annotated[
         str,
         typer.Option(
             "-R",
             "--retriever",
             help="Name of custom retriever to use",
-            click_type=click.Choice(["active_usings"]),
+            click_type=click.Choice(["active_usings", "language_docs"]),
         ),
     ] = None,
     max_tokens: Annotated[
@@ -430,6 +448,7 @@ def document(
         ),
     ] = None,
 ):
+    refiner_types = [REFINERS[r] for r in refiner_types]
     model_arguments = dict(temperature=temperature)
     collections_config = get_collections_config()
     kwargs = dict(
@@ -441,7 +460,7 @@ def document(
         db_path=db_loc,
         db_config=collections_config,
         splitter_type=splitter_type,
-        refiner_type=refiner_type,
+        refiner_types=refiner_types,
         retriever_type=retriever_type,
     )
     if doc_mode == "madlibs":
@@ -458,12 +477,6 @@ def document(
     documenter.translate(input_dir, output_dir, overwrite, collection)
 
 
-def get_subclasses(cls):
-    return set(cls.__subclasses__()).union(
-        set(s for c in cls.__subclasses__() for s in get_subclasses(c))
-    )
-
-
 @app.command()
 def aggregate(
     input_dir: Annotated[
@@ -578,6 +591,115 @@ def aggregate(
     aggregator.translate(input_dir, output_dir, overwrite, collection)
 
 
+@app.command(
+    help="Partition input code using an LLM.",
+    no_args_is_help=True,
+)
+def partition(
+    input_dir: Annotated[
+        Path,
+        typer.Option(
+            "--input",
+            "-i",
+            help="The directory containing the source code to be partitioned. ",
+        ),
+    ],
+    language: Annotated[
+        str,
+        typer.Option(
+            "--language",
+            "-l",
+            help="The language of the source code.",
+            click_type=click.Choice(sorted(LANGUAGES)),
+        ),
+    ],
+    output_dir: Annotated[
+        Path,
+        typer.Option(
+            "--output-dir", "-o", help="The directory to store the partitioned code in."
+        ),
+    ],
+    llm_name: Annotated[
+        str,
+        typer.Option(
+            "--llm",
+            "-L",
+            help="The custom name of the model set with 'janus llm add'.",
+        ),
+    ] = "gpt-4o",
+    max_prompts: Annotated[
+        int,
+        typer.Option(
+            "--max-prompts",
+            "-m",
+            help="The maximum number of times to prompt a model on one functional block "
+            "before exiting the application. This is to prevent wasting too much money.",
+        ),
+    ] = 10,
+    overwrite: Annotated[
+        bool,
+        typer.Option(
+            "--overwrite/--preserve",
+            help="Whether to overwrite existing files in the output directory",
+        ),
+    ] = False,
+    temperature: Annotated[
+        float,
+        typer.Option("--temperature", "-t", help="Sampling temperature.", min=0, max=2),
+    ] = 0.7,
+    splitter_type: Annotated[
+        str,
+        typer.Option(
+            "-S",
+            "--splitter",
+            help="Name of custom splitter to use",
+            click_type=click.Choice(list(CUSTOM_SPLITTERS.keys())),
+        ),
+    ] = "file",
+    refiner_types: Annotated[
+        list[str],
+        typer.Option(
+            "-r",
+            "--refiner",
+            help="List of refiner types to use. Add -r for each refiner to use in\
+                refinement chain",
+            click_type=click.Choice(list(REFINERS.keys())),
+        ),
+    ] = ["JanusRefiner"],
+    max_tokens: Annotated[
+        int,
+        typer.Option(
+            "--max-tokens",
+            "-M",
+            help="The maximum number of tokens the model will take in. "
+            "If unspecificed, model's default max will be used.",
+        ),
+    ] = None,
+    partition_token_limit: Annotated[
+        int,
+        typer.Option(
+            "--partition-tokens",
+            "-pt",
+            help="The limit on the number of tokens per partition.",
+        ),
+    ] = 8192,
+):
+    refiner_types = [REFINERS[r] for r in refiner_types]
+    model_arguments = dict(temperature=temperature)
+    kwargs = dict(
+        model=llm_name,
+        model_arguments=model_arguments,
+        source_language=language,
+        max_prompts=max_prompts,
+        max_tokens=max_tokens,
+        splitter_type=splitter_type,
+        refiner_types=refiner_types,
+        partition_token_limit=partition_token_limit,
+    )
+    partitioner = Partitioner(**kwargs)
+    partitioner.translate(input_dir, output_dir, overwrite)
+
+
 @app.command(
     help="Diagram input code using an LLM.",
     no_args_is_help=True,
@@ -667,25 +789,27 @@ def diagram(
             click_type=click.Choice(list(CUSTOM_SPLITTERS.keys())),
         ),
     ] = "file",
-    refiner_type: Annotated[
-        str,
+    refiner_types: Annotated[
+        list[str],
         typer.Option(
             "-r",
             "--refiner",
-            help="Name of custom refiner to use",
+            help="List of refiner types to use. Add -r for each refiner to use in\
+                refinement chain",
             click_type=click.Choice(list(REFINERS.keys())),
         ),
-    ] = "none",
+    ] = ["JanusRefiner"],
     retriever_type: Annotated[
         str,
         typer.Option(
             "-R",
             "--retriever",
             help="Name of custom retriever to use",
-            click_type=click.Choice(["active_usings"]),
+            click_type=click.Choice(["active_usings", "language_docs"]),
         ),
     ] = None,
 ):
+    refiner_types = [REFINERS[r] for r in refiner_types]
     model_arguments = dict(temperature=temperature)
     collections_config = get_collections_config()
     diagram_generator = DiagramGenerator(
@@ -696,7 +820,7 @@ def diagram(
         db_path=db_loc,
         db_config=collections_config,
         splitter_type=splitter_type,
-        refiner_type=refiner_type,
+        refiner_types=refiner_types,
         retriever_type=retriever_type,
         diagram_type=diagram_type,
         add_documentation=add_documentation,
@@ -704,6 +828,139 @@ def diagram(
     diagram_generator.translate(input_dir, output_dir, overwrite, collection)
 
 
+@app.command(
+    help="LLM self evaluation",
+    no_args_is_help=True,
+)
+def llm_self_eval(
+    input_dir: Annotated[
+        Path,
+        typer.Option(
+            "--input",
+            "-i",
+            help="The directory containing the source code to be evaluated. "
+            "The files should all be in one flat directory.",
+        ),
+    ],
+    language: Annotated[
+        str,
+        typer.Option(
+            "--language",
+            "-l",
+            help="The language of the source code.",
+            click_type=click.Choice(sorted(LANGUAGES)),
+        ),
+    ],
+    output_dir: Annotated[
+        Path,
+        typer.Option(
+            "--output-dir", "-o", help="The directory to store the evaluations in."
+        ),
+    ],
+    llm_name: Annotated[
+        str,
+        typer.Option(
+            "--llm",
+            "-L",
+            help="The custom name of the model set with 'janus llm add'.",
+        ),
+    ] = "gpt-4o",
+    evaluation_type: Annotated[
+        str,
+        typer.Option(
+            "--evaluation-type",
+            "-e",
+            help="Type of output to evaluate.",
+            click_type=click.Choice(["incose", "comments"]),
+        ),
+    ] = "incose",
+    max_prompts: Annotated[
+        int,
+        typer.Option(
+            "--max-prompts",
+            "-m",
+            help="The maximum number of times to prompt a model on one functional block "
+            "before exiting the application. This is to prevent wasting too much money.",
+        ),
+    ] = 10,
+    overwrite: Annotated[
+        bool,
+        typer.Option(
+            "--overwrite/--preserve",
+            help="Whether to overwrite existing files in the output directory",
+        ),
+    ] = False,
+    temperature: Annotated[
+        float,
+        typer.Option("--temperature", "-t", help="Sampling temperature.", min=0, max=2),
+    ] = 0.7,
+    collection: Annotated[
+        str,
+        typer.Option(
+            "--collection",
+            "-c",
+            help="If set, will put the translated result into a Chroma DB "
+            "collection with the name provided.",
+        ),
+    ] = None,
+    splitter_type: Annotated[
+        str,
+        typer.Option(
+            "-S",
+            "--splitter",
+            help="Name of custom splitter to use",
+            click_type=click.Choice(list(CUSTOM_SPLITTERS.keys())),
+        ),
+    ] = "file",
+    refiner_types: Annotated[
+        list[str],
+        typer.Option(
+            "-r",
+            "--refiner",
+            help="List of refiner types to use. Add -r for each refiner to use in\
+                refinement chain",
+            click_type=click.Choice(list(REFINERS.keys())),
+        ),
+    ] = ["JanusRefiner"],
+    eval_items_per_request: Annotated[
+        int,
+        typer.Option(
+            "--eval-items-per-request",
+            "-rc",
+            help="The maximum number of evaluation items per request",
+        ),
+    ] = None,
+    max_tokens: Annotated[
+        int,
+        typer.Option(
+            "--max-tokens",
+            "-M",
+            help="The maximum number of tokens the model will take in. "
+            "If unspecificed, model's default max will be used.",
+        ),
+    ] = None,
+):
+    model_arguments = dict(temperature=temperature)
+    refiner_types = [REFINERS[r] for r in refiner_types]
+    kwargs = dict(
+        eval_items_per_request=eval_items_per_request,
+        model=llm_name,
+        model_arguments=model_arguments,
+        source_language=language,
+        max_prompts=max_prompts,
+        max_tokens=max_tokens,
+        splitter_type=splitter_type,
+        refiner_types=refiner_types,
+    )
+    # Setting parser type here
+    if evaluation_type == "incose":
+        evaluator = RequirementEvaluator(**kwargs)
+    elif evaluation_type == "comments":
+        evaluator = InlineCommentEvaluator(**kwargs)
+
+    evaluator.translate(input_dir, output_dir, overwrite, collection)
+
+
 @db.command("init", help="Connect to or create a database.")
 def db_init(
     path: Annotated[
@@ -1005,13 +1262,12 @@ def llm_add(
             show_choices=False,
         )
         params = dict(
-            # OpenAI uses the "model_name" key for what we're calling "long_model_id"
-            model_name=MODEL_ID_TO_LONG_ID[model_id],
+            model_name=model_name,
             temperature=0.7,
             n=1,
         )
-        max_tokens = TOKEN_LIMITS[MODEL_ID_TO_LONG_ID[model_id]]
-        model_cost = COST_PER_1K_TOKENS[MODEL_ID_TO_LONG_ID[model_id]]
+        max_tokens = TOKEN_LIMITS[model_name]
+        model_cost = COST_PER_1K_TOKENS[model_name]
         cfg = {
             "model_type": model_type,
             "model_id": model_id,

{janus_llm-4.1.0 → janus_llm-4.3.1}/janus/converter/__init__.py

@@ -2,5 +2,6 @@ from janus.converter.converter import Converter
 from janus.converter.diagram import DiagramGenerator
 from janus.converter.document import Documenter, MadLibsDocumenter, MultiDocumenter
 from janus.converter.evaluate import Evaluator
+from janus.converter.partition import Partitioner
 from janus.converter.requirements import RequirementsDocumenter
 from janus.converter.translate import Translator