janus-llm 3.5.3__py3-none-any.whl → 4.0.0__py3-none-any.whl

janus/metrics/metric.py

@@ -8,6 +8,7 @@ import typer
 from typing_extensions import Annotated
 
 from janus.llm import load_model
+from janus.llm.model_callbacks import COST_PER_1K_TOKENS
 from janus.metrics.cli import evaluate
 from janus.metrics.file_pairing import FILE_PAIRING_METHODS
 from janus.metrics.splitting import SPLITTING_METHODS
@@ -135,7 +136,7 @@ def metric(
                 **kwargs,
             ):
                 out = []
-                llm, _, token_limit, model_cost = load_model(llm_name)
+                llm = load_model(llm_name)
                 if json_file_name is not None:
                     with open(json_file_name, "r") as f:
                         json_obj = json.load(f)
@@ -171,8 +172,8 @@ def metric(
                         out_file=out_file,
                         lang=language,
                         llm=llm,
-                        token_limit=token_limit,
-                        model_cost=model_cost,
+                        token_limit=llm.token_limit,
+                        model_cost=COST_PER_1K_TOKENS[llm.model_id],
                     )
                 else:
                     raise ValueError(
@@ -187,8 +188,8 @@ def metric(
                             progress,
                             language,
                             llm,
-                            token_limit,
-                            model_cost,
+                            llm.token_limit,
+                            COST_PER_1K_TOKENS[llm.model_id],
                             *args,
                             **kwargs,
                         )
@@ -199,8 +200,8 @@ def metric(
                         progress,
                         language,
                         llm,
-                        token_limit,
-                        model_cost,
+                        llm.token_limit,
+                        COST_PER_1K_TOKENS[llm.model_id],
                         *args,
                         **kwargs,
                     )
@@ -296,7 +297,7 @@ def metric(
                 *args,
                 **kwargs,
             ):
-                llm, _, token_limit, model_cost = load_model(llm_name)
+                llm = load_model(llm_name)
                 if json_file_name is not None:
                     with open(json_file_name, "r") as f:
                         json_obj = json.load(f)
@@ -328,8 +329,8 @@ def metric(
                         out_file=out_file,
                         lang=language,
                         llm=llm,
-                        token_limit=token_limit,
-                        model_cost=model_cost,
+                        token_limit=llm.token_limit,
+                        model_cost=COST_PER_1K_TOKENS[llm.model_id],
                     )
                 else:
                     raise ValueError(
@@ -344,8 +345,8 @@ def metric(
                             progress,
                             language,
                             llm,
-                            token_limit,
-                            model_cost,
+                            llm.token_limit,
+                            COST_PER_1K_TOKENS[llm.model_id],
                             *args,
                             **kwargs,
                         )
@@ -356,8 +357,8 @@ def metric(
                         progress,
                         language,
                         llm,
-                        token_limit,
-                        model_cost,
+                        llm.token_limit,
+                        COST_PER_1K_TOKENS[llm.model_id],
                         *args,
                         **kwargs,
                     )

janus/parsers/uml.py

@@ -1,7 +1,7 @@
 import re
 import subprocess  # nosec
 from pathlib import Path
-from typing import List, Tuple
+from tempfile import NamedTemporaryFile
 
 from langchain_core.exceptions import OutputParserException
 from langchain_core.messages import BaseMessage
@@ -13,39 +13,76 @@ log = create_logger(__name__)
 
 
 class UMLSyntaxParser(CodeParser):
-    def _get_uml_output(self, file: Path) -> Tuple[str, str]:
-        # NOTE: running subprocess with shell=False, added nosec to label that we know
-        # risk exists
+    def _check_plantuml(self, text: str) -> None:
+        # Leading newlines can break the parser, remove them
+        text = text.replace("\\n", "\n").strip()
+
+        # Write the text to a temporary file (automatically deleted)
+        file = NamedTemporaryFile()
+        fname = file.name
+        with open(fname, "w") as fin:
+            fin.write(text)
+
         try:
             plantuml_path = Path.home().expanduser() / ".janus/lib/plantuml.jar"
+            # NOTE: running subprocess with shell=False, added nosec to
+            # label that we know risk exists
             res = subprocess.run(
-                ["java", "-jar", plantuml_path, file],
+                ["java", "-jar", plantuml_path, fname],
                 stdout=subprocess.PIPE,
                 stderr=subprocess.PIPE,
             )  # nosec
             stdout = res.stdout.decode("utf-8")
             stderr = res.stderr.decode("utf-8")
         except FileNotFoundError:
-            log.warning("Plant UML executable not found, skipping syntax check")
-            stdout = ""
-            stderr = ""
-        return stdout, stderr
+            err_txt = (
+                "Plant UML executable not found. Either choose a different parser"
+                " or install with `bash scripts/install_plantuml.sh`. Java and"
+                " graphviz are dependencies for the tool, they must also be installed."
+            )
+            log.error(err_txt)
+            raise Exception(err_txt)
+
+        # Check for bad outputs, raise OutputParserExceptions if so
+        if "Error" in stderr or "Error" in stdout:
+            err_txt = "Recieved UML parsing error(s)."
+
+            line_nos = self._get_error_lines(stderr) + self._get_error_lines(stdout)
+            lines = text.split("\n")
+            for i in line_nos:
+                i0 = max(0, i - 3)
+                i1 = min(len(lines) - 1, i + 2)
+                err_lines = [
+                    f"> {lines[j]}" if j == i - 1 else f"  {lines[j]}"
+                    for j in range(i0, i1)
+                ]
+                if i0:
+                    err_lines.insert(0, "  ...")
+                if i1 < (len(lines) - 1):
+                    err_lines.append("  ...")
 
-    def _get_errs(self, s: str) -> List[str]:
-        return [x.group() for x in re.finditer(r"Error (.*)\n", s)]
+                err_txt += f"\nError located at line {i} must be fixed:\n"
+                err_txt += "\n".join(err_lines)
+            log.warning(err_txt)
+            raise OutputParserException(err_txt)
+
+        if "Warning" in stdout or "Warning" in stderr:
+            err_txt = "Recieved UML parsing warning (often due to missing PLANTUML)."
+            if stderr:
+                err_txt += f"\nSTDERR:\n```\n{stderr.strip()}\n```\n"
+            if stdout:
+                err_txt += f"\nSTDOUT:\n```\n{stdout.strip()}\n```\n"
+
+            log.warning(err_txt)
+            raise OutputParserException(err_txt)
+
+    def _get_error_lines(self, s: str) -> list[int]:
+        return [int(x.group(1)) for x in re.finditer(r"Error line (\d+) in file:", s)]
+
+    def _get_warns(self, s: str) -> list[str]:
+        return [x.group() for x in re.finditer(r"Warning: (.*)\n", s)]
 
     def parse(self, text: str | BaseMessage) -> str:
         text = super().parse(text)
-        janus_path = Path.home().expanduser() / Path(".janus")
-        if not janus_path.exists():
-            janus_path.mkdir()
-        temp_file_path = janus_path / "tmp.txt"
-        with open(temp_file_path, "w") as f:
-            f.write(text)
-        uml_std_out, uml_std_err = self._get_uml_output(temp_file_path)
-        uml_errs = self._get_errs(uml_std_out) + self._get_errs(uml_std_err)
-        if len(uml_errs) > 0:
-            raise OutputParserException(
-                "Error: Received UML Errors:\n" + "\n".join(uml_errs)
-            )
+        self._check_plantuml(text)
         return text

janus/refiners/refiner.py

@@ -1,73 +1,115 @@
-from langchain_core.prompts import ChatPromptTemplate
+from typing import Any
 
-from janus.llm.models_info import MODEL_PROMPT_ENGINES
+from langchain.output_parsers import RetryWithErrorOutputParser
+from langchain_core.output_parsers import StrOutputParser
+from langchain_core.prompt_values import PromptValue
+from langchain_core.runnables import RunnableSerializable
 
+from janus.llm.models_info import MODEL_PROMPT_ENGINES, JanusModel
+from janus.parsers.parser import JanusParser
+from janus.utils.logger import create_logger
 
-class Refiner:
-    def refine(
-        self,
-        original_prompt: str,
-        previous_prompt: str,
-        previous_output: str,
-        errors: str,
-        **kwargs,
-    ) -> tuple[ChatPromptTemplate, dict[str, str]]:
-        """Creates a new prompt based on feedback from original results
-
-        Arguments:
-            original_prompt: original prompt used to produce output
-            original_output: origial output of llm
-            errors: list of errors detected by parser
-
-        Returns:
-            Tuple of new prompt and prompt arguments
-        """
+log = create_logger(__name__)
+
+
+class JanusRefiner(JanusParser):
+    parser: JanusParser
+
+    def parse_runnable(self, input: dict[str, Any]) -> Any:
+        return self.parse_completion(**input)
+
+    def parse_completion(self, completion: str, **kwargs) -> Any:
+        return self.parser.parse(completion)
+
+    def parse(self, text: str) -> str:
         raise NotImplementedError
 
 
-class BasicRefiner(Refiner):
+class FixParserExceptions(JanusRefiner, RetryWithErrorOutputParser):
+    def __init__(self, llm: JanusModel, parser: JanusParser, max_retries: int):
+        retry_prompt = MODEL_PROMPT_ENGINES[llm.model_id](
+            source_language="text",
+            prompt_template="refinement/fix_exceptions",
+        ).prompt
+        chain = retry_prompt | llm | StrOutputParser()
+        RetryWithErrorOutputParser.__init__(
+            self, parser=parser, retry_chain=chain, max_retries=max_retries
+        )
+
+    def parse_completion(
+        self, completion: str, prompt_value: PromptValue, **kwargs
+    ) -> Any:
+        return self.parse_with_prompt(completion, prompt_value=prompt_value)
+
+
+class ReflectionRefiner(JanusRefiner):
+    max_retries: int
+    reflection_chain: RunnableSerializable
+    revision_chain: RunnableSerializable
+
     def __init__(
         self,
-        prompt_name: str,
-        model_id: str,
-        source_language: str,
-    ) -> None:
-        """Basic refiner, asks llm to fix output of previous prompt given errors
-
-        Arguments:
-            prompt_name: refinement prompt name to use
-            model_id: ID of the llm to use. Found in models_info.py
-            source_language: source_langauge to use
-        """
-        self._prompt_name = prompt_name
-        self._model_id = model_id
-        self._source_language = source_language
-
-    def refine(
-        self,
-        original_prompt: str,
-        previous_prompt: str,
-        previous_output: str,
-        errors: str,
-        **kwargs,
-    ) -> tuple[ChatPromptTemplate, dict[str, str]]:
-        """Creates a new prompt based on feedback from original results
-
-        Arguments:
-            original_prompt: original prompt used to produce output
-            original_output: origial output of llm
-            errors: list of errors detected by parser
-
-        Returns:
-            Tuple of new prompt and prompt arguments
-        """
-        prompt_engine = MODEL_PROMPT_ENGINES[self._model_id](
-            prompt_template=self._prompt_name,
-            source_language=self._source_language,
+        llm: JanusModel,
+        parser: JanusParser,
+        max_retries: int,
+        prompt_template_name: str = "refinement/reflection",
+    ):
+        reflection_prompt = MODEL_PROMPT_ENGINES[llm.model_id](
+            source_language="text",
+            prompt_template=prompt_template_name,
+        ).prompt
+        revision_prompt = MODEL_PROMPT_ENGINES[llm.model_id](
+            source_language="text",
+            prompt_template="refinement/revision",
+        ).prompt
+
+        reflection_chain = reflection_prompt | llm | StrOutputParser()
+        revision_chain = revision_prompt | llm | StrOutputParser()
+        super().__init__(
+            reflection_chain=reflection_chain,
+            revision_chain=revision_chain,
+            parser=parser,
+            max_retries=max_retries,
+        )
+
+    def parse_completion(
+        self, completion: str, prompt_value: PromptValue, **kwargs
+    ) -> Any:
+        for retry_number in range(self.max_retries):
+            reflection = self.reflection_chain.invoke(
+                dict(
+                    prompt=prompt_value.to_string(),
+                    completion=completion,
+                )
+            )
+            if reflection.strip() == "LGTM":
+                return self.parser.parse(completion)
+            if not retry_number:
+                log.info(f"Completion:\n{completion}")
+            log.info(f"Reflection:\n{reflection}")
+            completion = self.revision_chain.invoke(
+                dict(
+                    prompt=prompt_value.to_string(),
+                    completion=completion,
+                    reflection=reflection,
+                )
+            )
+            log.info(f"Revision:\n{completion}")
+
+        return self.parser.parse(completion)
+
+
+class HallucinationRefiner(ReflectionRefiner):
+    def __init__(self, **kwargs):
+        super().__init__(
+            prompt_template_name="refinement/hallucination",
+            **kwargs,
         )
-        prompt_arguments = {
-            "ORIGINAL_PROMPT": original_prompt,
-            "OUTPUT": previous_output,
-            "ERRORS": errors,
-        }
-        return prompt_engine.prompt, prompt_arguments
+
+
+REFINERS = dict(
+    none=JanusRefiner,
+    parser=FixParserExceptions,
+    reflection=ReflectionRefiner,
+    hallucination=HallucinationRefiner,
+)

janus/retrievers/retriever.py

@@ -0,0 +1,42 @@
+from langchain_core.retrievers import BaseRetriever
+from langchain_core.runnables import Runnable, RunnableConfig
+
+from janus.language.block import CodeBlock
+
+
+class JanusRetriever(Runnable):
+    def __init__(self) -> None:
+        super().__init__()
+
+    def invoke(
+        self, input: CodeBlock, config: RunnableConfig | None = None, **kwargs
+    ) -> dict:
+        kwargs.update(context=self.get_context(input))
+        return kwargs
+
+    def get_context(self, code_block: CodeBlock) -> str:
+        return ""
+
+
+class ActiveUsingsRetriever(JanusRetriever):
+    def get_context(self, code_block: CodeBlock) -> str:
+        context = "\n".join(
+            f"{context_tag}: {context}"
+            for context_tag, context in code_block.context_tags.items()
+        )
+        return f"You may use the following additional context: {context}"
+
+
+class TextSearchRetriever(JanusRetriever):
+    retriever: BaseRetriever
+
+    def __init__(self, retriever: BaseRetriever):
+        super().__init__()
+        self.retriever = retriever
+
+    def get_context(self, code_block: CodeBlock) -> str:
+        if code_block.text is None:
+            return ""
+        docs = self.retriever.invoke(code_block.text)
+        context = "\n\n".join(doc.page_content for doc in docs)
+        return f"You may use the following additional context: {context}"

{janus_llm-3.5.3.dist-info → janus_llm-4.0.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: janus-llm
-Version: 3.5.3
+Version: 4.0.0
 Summary: A transcoding library using LLMs.
 Home-page: https://github.com/janus-llm/janus-llm
 License: Apache 2.0

{janus_llm-3.5.3.dist-info → janus_llm-4.0.0.dist-info}/RECORD

@@ -1,19 +1,19 @@
-janus/__init__.py,sha256=7OGST6n7Xx5q_ty0BF_UaZ2cjaepl2HSiPRJXY7XZcQ,361
+janus/__init__.py,sha256=gL7W9Io7zembI938nQAbVjAY107vRS0acvWfDDrEVyg,361
 janus/__main__.py,sha256=lEkpNtLVPtFo8ySDZeXJ_NXDHb0GVdZFPWB4gD4RPS8,64
 janus/_tests/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 janus/_tests/conftest.py,sha256=V7uW-oq3YbFiRPvrq15YoVVrA1n_83pjgiyTZ-IUGW8,963
 janus/_tests/test_cli.py,sha256=6ef7h11bg4i7Q6L1-r0ZdcY7YrH4n472kvDiA03T4c8,4275
-janus/cli.py,sha256=RuXaM9rWKW4bvy4ML_epcygDIMSAERM9Os7giO-Ivbk,37810
+janus/cli.py,sha256=lX8GzLOXan8nqjy_dx-3EF4qaLOVTsyDqp0OLBZctw0,38177
 janus/converter/__init__.py,sha256=U2EOMcCykiC0ZqhorNefOP_04hOF18qhYoPKrVp1Vrk,345
 janus/converter/_tests/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 janus/converter/_tests/test_translate.py,sha256=yzcFEGc_z8QmBBBmC9dZnfL9tT8CD1rkpc8Hz44Jp4c,5631
 janus/converter/aggregator.py,sha256=MuAXMKmq6PuUo_w6ljyiuDn81Gk2dN-Ci7FVeLc6vhs,1966
-janus/converter/converter.py,sha256=GL4moCOi5vvvsEjAf_7Vrn5GHq12nptH47AAP4soeQc,27956
-janus/converter/diagram.py,sha256=sqztG_8T_gQ4njk5V7tg1QCgT6gFbL3CUy7t2l0X5i4,5283
+janus/converter/converter.py,sha256=HWw-a4j-Qq0AWDyyk6cMVOwYjJeXxa3bvs9OOlKo_KI,25965
+janus/converter/diagram.py,sha256=-wktVBPrSBgNIQfHIfa2bJNg6L9CYJQgrr9-xU8DFPw,1646
 janus/converter/document.py,sha256=qNt2UncMheUBadXCFHGq74tqCrvZub5DCgZpd3Qa54o,4564
 janus/converter/evaluate.py,sha256=APWQUY3gjAXqkJkPzvj0UA4wPK3Cv9QSJLM-YK9t-ng,476
 janus/converter/requirements.py,sha256=9tvQ40FZJtG8niIFn45gPQCgKKHVPPoFLinBv6RAqO4,2027
-janus/converter/translate.py,sha256=0brQTlSfBYmXtoM8QYIOiyr0LrTr0S1n68Du-BR7_WQ,4236
+janus/converter/translate.py,sha256=S1DPZdmX9Vrn_sJPcobvXmhmS8U53yl5cRXjsmXPtas,4246
 janus/embedding/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 janus/embedding/_tests/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 janus/embedding/_tests/test_collections.py,sha256=llg-JSuRRFhKkHFiWWSHEWV3iaT6Lwue0lp2tEml9io,2668
@@ -29,20 +29,20 @@ janus/language/_tests/test_combine.py,sha256=sjVVPUg4LYkAmazXGUw_S1xPrzWm67_0tCx
 janus/language/_tests/test_splitter.py,sha256=Hqexa39LLEXlK3ZUw7Zot4PUIACvye2vkq0Jaox0T10,373
 janus/language/alc/__init__.py,sha256=j7vOMGhT1Vri6p8dsjSaY-fkO5uFn0sJ0nrNGGvcizM,42
 janus/language/alc/_tests/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-janus/language/alc/_tests/test_alc.py,sha256=NgVeOctm9zf-S328DdUNn9et_-lK1t5O0O2FKElb91Q,1027
-janus/language/alc/alc.py,sha256=hiHpS-odevqbOS6TkJJgghiGXKU1ObwrNzAI66DNsP8,6598
+janus/language/alc/_tests/test_alc.py,sha256=jrvZCsz3uvbq6pqTKmymWNDgb20HHs69tj9TRbx69aM,1018
+janus/language/alc/alc.py,sha256=fKZDtbeLXiJ2e8t-eWSjeAW6WCMCTn2Fw7_jxvMPdNc,6863
 janus/language/binary/__init__.py,sha256=AlNAe12ZA366kcGSrQ1FJyOdbwxFqGBFkYR2K6yL818,51
 janus/language/binary/_tests/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-janus/language/binary/_tests/test_binary.py,sha256=SDdI6tsQj9yXle7wBsksHuKULLMHv7mNgUkDx1nCvpw,1733
-janus/language/binary/binary.py,sha256=jcc-LZx8Ss-g4j0a691U2yVJV6JGV_zpE_6y_aDq4Cw,6579
+janus/language/binary/_tests/test_binary.py,sha256=cIKIxjj6kIY3rcxLwqUPESP9bxWrHqMHx9TNuICgfeQ,1724
+janus/language/binary/binary.py,sha256=PHsVa8jcM7sW9aTboGRWXj6ewQznz0kVPNWtP4B9YPU,6555
 janus/language/binary/reveng/decompile_script.py,sha256=veW51oJzuO-4UD3Er062jXZ_FYtTFo9OCkl82Z2xr6A,2182
 janus/language/block.py,sha256=2rjAYUosHFfWRgLnzf50uAgTMST4Md9Kx6JrlUfEfX4,9398
 janus/language/combine.py,sha256=Wtve06fa-_Wjv_V5RIf1Nfmg0UxcOEtFNj4vVHpSNbo,2940
 janus/language/file.py,sha256=jy-cReAoI6F97TXR5bbhPyt8XyUZCdFYnVboubDA_y4,571
 janus/language/mumps/__init__.py,sha256=-Ou_wJ-JgHezfp1dub2_qCYNiK9wO-zo2MlqxM9qiwE,48
 janus/language/mumps/_tests/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-janus/language/mumps/_tests/test_mumps.py,sha256=dl-87kuWUFu_sZZl-vZYgKB-HTQjQhDuxU1cT2hTAX4,1019
-janus/language/mumps/mumps.py,sha256=gBxk6Dhf4HqhU6ymweQuZqwEp-AME2t2JHjmzPNKIqc,7406
+janus/language/mumps/_tests/test_mumps.py,sha256=YxnQkobM9VR_Zow2oLaZxZs8bPipvbFWSza4zlWbMi8,1010
+janus/language/mumps/mumps.py,sha256=-eyafFN9IeUQMU7yEsICJDrDb5tkYV_EydqFCjcuM_o,7381
 janus/language/mumps/patterns.py,sha256=fLbgEnPYYggfqTnBwXf0XbnlU3g_bM4tsmO1YHYv5Jw,2313
 janus/language/naive/__init__.py,sha256=_Gq4inONyVYxe8WLB59d_69kqGbtF40BGKoJPnK4tZ0,258
 janus/language/naive/basic_splitter.py,sha256=RM9pJK2YkHfb6_EFEV-dh_rLqkjS6v0cn3ASPf8A6Fg,459
@@ -51,14 +51,14 @@ janus/language/naive/registry.py,sha256=8YQX1q0IdAm7t69-oC_00I-vfkdRnHuX-OD3KEjE
 janus/language/naive/simple_ast.py,sha256=YzeUJomVsnttJc8tI9eDROb2Hx9Vm9XKmOnLEp3TkzI,3112
 janus/language/naive/tag_splitter.py,sha256=IXWMn9tBVUGAtzvQi89GhoZ6g7fPXk5MzO0kMCr2mb0,2045
 janus/language/node.py,sha256=baoYFtapwBQqBtUN6EvHFYRkbR-EcEw1b3fQvH9zIAM,204
-janus/language/splitter.py,sha256=ckrNgN2G3c2cp7AJeZmem21H2FF2pYZtOivqwyLIb6M,17029
+janus/language/splitter.py,sha256=pYvDhGAYDDP7E4CZeNn76I6zVnHMNj6gTFUegKXyRPk,17005
 janus/language/treesitter/__init__.py,sha256=mUliw7ZJLZ8NkJKyUQMSoUV82hYXE0HvLHrEdGPJF4Q,43
 janus/language/treesitter/_tests/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-janus/language/treesitter/_tests/test_treesitter.py,sha256=4S_UdH6AfJ0j6hyInZ2CFLRqLXb0Bq-AMYd7A8-Y3iU,2188
-janus/language/treesitter/treesitter.py,sha256=FU86H8j2cfRLtwyNeEnf9A6gzZEvDwmnFCIrQymUJso,7541
+janus/language/treesitter/_tests/test_treesitter.py,sha256=fmr_mFSja7vaCVu0TVyLDua3A94jMjY4AqSC5NqnOdQ,2179
+janus/language/treesitter/treesitter.py,sha256=q7fqfFxt7QsqM6tE39uqutRMsOfEgBd3omv7zVZSEOc,7517
 janus/llm/__init__.py,sha256=TKLYvnsWKWfxMucy-lCLQ-4bkN9ENotJZDywDEQmrKg,45
-janus/llm/model_callbacks.py,sha256=K7P5NY-rf7IYRAFHnZ3kzhrQWE6g_najx8uxlaSiz3E,7110
-janus/llm/models_info.py,sha256=mSY0lnMN7hg5-Gtbz207-Y4EZiCzB8GDU9KV2jMCKYE,8668
+janus/llm/model_callbacks.py,sha256=iaLsnHt2LcMgHPW4ouQdot9ahOHUOBCtydpVaag1Zak,7564
+janus/llm/models_info.py,sha256=GxAr7iwjcvM7RWUe1I6AuOqiYA-Wnc4Mr8DEVBq91KA,9662
 janus/metrics/__init__.py,sha256=AsxtZJUzZiXJPr2ehPPltuYP-ddechjg6X85WZUO7mA,241
 janus/metrics/_tests/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 janus/metrics/_tests/reference.py,sha256=hiaJPP9CXkvFBV_wL-gOe_BzELTw0nvB6uCxhxtIiE8,13
@@ -77,7 +77,7 @@ janus/metrics/cli.py,sha256=Duuw2RF47Z-t1pal0cg3L_-N_91rx29krirqtIwjYLY,157
 janus/metrics/complexity_metrics.py,sha256=9sHxa9tY9IgdFC4grCmDSuu8Nkphm_FWbTecxgW78go,6574
 janus/metrics/file_pairing.py,sha256=b0848gbDTqKVge-uGiujtHdsgcbkpJ04nWF2ul_gbCw,3749
 janus/metrics/llm_metrics.py,sha256=nzzJMSnoZfgtzzltjoCu1l4-Qwuycd-V3z17AtStYdY,5700
-janus/metrics/metric.py,sha256=nZJQY-_AdiRFz6Af38t6TOam1I_yCns89-aOHwPRIos,17011
+janus/metrics/metric.py,sha256=ZQXz2LLXA3_WW6BluF8Q2LdJKYVWZLYmhO4Ai116ZPY,17168
 janus/metrics/reading.py,sha256=RyOSebENWPGD1m_-QfKM9v2vfNAQVkiZ2Oyynb2HrVI,3253
 janus/metrics/rouge_score.py,sha256=XnA2ABAceIzzA_9rIYDaMtyD4pphHxXVwxaYvMcqZWE,2878
 janus/metrics/similarity.py,sha256=NKd-m8hMB27i1_1CQkG5o7HrAK2wTr89B21Y36kqHiE,1630
@@ -89,12 +89,12 @@ janus/parsers/code_parser.py,sha256=3l0HfzgrvJuiwk779s9ZsgUl3xbp1nE1qZxh8aDYRBI,
 janus/parsers/doc_parser.py,sha256=0pUsNZ9hKQLjIi8L8BgkOBHQZ_EGoFLHrBQ4hoDkjSw,5862
 janus/parsers/eval_parser.py,sha256=Gjh6aTZgpYd2ASJUEPMo4LpCL00cBmbOqc4KM3hy8x8,2922
 janus/parsers/parser.py,sha256=y6VV64bgVidf-oEFla3I--_28tnJsPBc6QUD_SkbfSE,1614
-janus/parsers/refiner_parser.py,sha256=W7ecDX7j6N-uWtRBYFD1EiEzDpDQws16nrVFzyArzf8,1632
 janus/parsers/reqs_parser.py,sha256=uRQC41Iqp22GjIvakb5UKv70UWHkcOTbOVl_RDnipYw,2438
-janus/parsers/uml.py,sha256=T9MzYbXz_gjlfgJlIMGSzgACcXLhoN3RnV9Y-5vYJ9A,1888
+janus/parsers/uml.py,sha256=SwaoG9QrHKQP8rSxlf3qu_rp7OMQqYSmLgDYBapOa9M,3379
 janus/prompts/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 janus/prompts/prompt.py,sha256=3796YXIzzIec9b0iUzd8VZlq-AdQbzq8qUGXLy4KH-0,10586
-janus/refiners/refiner.py,sha256=GkV4oUSCrLAhyDJY2aY_Jt8PRF3sC6-bv58YbL2PaNk,2227
+janus/refiners/refiner.py,sha256=lKAVS9Nszx8aO7hLB0xzdMmUNfhX7K5SaXwcit2QBHs,3774
+janus/retrievers/retriever.py,sha256=WzsYn329GbyJfmJ9KyU9yiYdhLkuS6WPKTKoHx5fBYQ,1339
 janus/utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 janus/utils/_tests/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 janus/utils/_tests/test_logger.py,sha256=jkkvrCTKwsFCsZtmyuvc-WJ0rC7LJi2Z91sIe4IiKzA,2209
@@ -102,8 +102,8 @@ janus/utils/_tests/test_progress.py,sha256=Rs_u5PiGjP-L-o6C1fhwfE1ig8jYu9Xo9s4p8
 janus/utils/enums.py,sha256=AoilbdiYyMvY2Mp0AM4xlbLSELfut2XMwhIM1S_msP4,27610
 janus/utils/logger.py,sha256=KZeuaMAnlSZCsj4yL0P6N-JzZwpxXygzACWfdZFeuek,2337
 janus/utils/progress.py,sha256=PIpcQec7SrhsfqB25LHj2CDDkfm9umZx90d9LZnAx6k,1469
-janus_llm-3.5.3.dist-info/LICENSE,sha256=_j0st0a-HB6MRbP3_BW3PUqpS16v54luyy-1zVyl8NU,10789
-janus_llm-3.5.3.dist-info/METADATA,sha256=tNfk-v8PG8wfl1Fj1MIa3vWJcHNZkxmUIOKsbvb_5WQ,4184
-janus_llm-3.5.3.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
-janus_llm-3.5.3.dist-info/entry_points.txt,sha256=OGhQwzj6pvXp79B0SaBD5apGekCu7Dwe9fZZT_TZ544,39
-janus_llm-3.5.3.dist-info/RECORD,,
+janus_llm-4.0.0.dist-info/LICENSE,sha256=_j0st0a-HB6MRbP3_BW3PUqpS16v54luyy-1zVyl8NU,10789
+janus_llm-4.0.0.dist-info/METADATA,sha256=LiMK2QiE9nzhufdK43eDYll_CbSThm2Wjit7bpG3XdE,4184
+janus_llm-4.0.0.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
+janus_llm-4.0.0.dist-info/entry_points.txt,sha256=OGhQwzj6pvXp79B0SaBD5apGekCu7Dwe9fZZT_TZ544,39
+janus_llm-4.0.0.dist-info/RECORD,,

janus/parsers/refiner_parser.py

@@ -1,46 +0,0 @@
-from langchain_core.exceptions import OutputParserException
-from langchain_core.language_models import BaseLanguageModel
-from langchain_core.messages import BaseMessage
-from langchain_core.output_parsers import BaseOutputParser
-
-from janus.parsers.parser import JanusParser
-from janus.refiners.refiner import Refiner
-
-
-class RefinerParser(JanusParser):
-    """Parser for performing refinement with a refiner
-
-    Properties:
-        llm: the language model to use
-        parser: the parser to use for parsing llm output
-        initial_prompt: initial prompt used to generate output
-        refiner: refiner that gives new subsequent prompts
-        max_retires: maximum number of times to attempt refining
-    """
-
-    class Config:
-        arbitrary_types_allowed = True
-
-    llm: BaseLanguageModel
-    parser: BaseOutputParser
-    initial_prompt: str
-    refiner: Refiner
-    max_retries: int
-
-    def parse(self, text: str | BaseMessage) -> str:
-        last_prompt = self.initial_prompt
-        for _ in range(self.max_retries):
-            try:
-                return self.parser.parse(text)
-            except OutputParserException as oe:
-                err = str(oe)
-                new_prompt, prompt_arguments = self.refiner.refine(
-                    self.initial_prompt, last_prompt, text, err
-                )
-                new_chain = new_prompt | self.llm
-                text = new_chain.invoke(prompt_arguments)
-                last_prompt = new_prompt.format(**prompt_arguments)
-
-        raise OutputParserException(
-            f"Error: unable to correct output after {self.max_retries} attempts"
-        )