Improve context reviewer prompter using CO-STAR. (#41)

* Improve context reviewer prompt by CO-STAR. * Minor refactor for check_format.
zh-plus · Jun 5, 2024 · a8450ec · a8450ec
1 parent 1868c6c
commit a8450ec
Show file tree

Hide file tree

Showing 5 changed files with 109 additions and 99 deletions.
diff --git a/README.md b/README.md
@@ -191,6 +191,7 @@ and [Anthropic](https://docs.anthropic.com/claude/docs/models-overview#model-com
 | `gpt-3.5-turbo`            | 0.5, 1.5                                        | 0.01                             |
 | `gpt-4-0125-preview`       | 10, 30                                          | 0.5                              |
 | `gpt-4-turbo-preview`      | 10, 30                                          | 0.5                              |
+| `gpt-4o`                   | 5, 15                                           | 0.25                             |
 | `claude-3-haiku-20240307`  | 0.25, 1.25                                      | 0.015                            |
 | `claude-3-sonnet-20240229` | 3, 15                                           | 0.2                              |
 | `claude-3-opus-20240229`   | 15, 75                                          | 1                                |

diff --git a/openlrc/agents.py b/openlrc/agents.py
@@ -7,8 +7,8 @@
 from openlrc.chatbot import route_chatbot
 from openlrc.context import TranslationContext, TranslateInfo
 from openlrc.logger import logger
-from openlrc.prompter import BaseTranslatePrompter, ContextReviewPrompter, potential_prefix_combo, \
-    ProofreaderPrompter, proofread_prefix
+from openlrc.prompter import BaseTranslatePrompter, ContextReviewPrompter, POTENTIAL_PREFIX_COMBOS, \
+    ProofreaderPrompter, PROOFREAD_PREFIX
 
 
 class Agent(abc.ABC):
@@ -70,7 +70,7 @@ def _extract_tag_content(self, content: str, tag: str) -> str:
         return match.group(1) if match else ''
 
     def _extract_translations(self, content: str) -> List[str]:
-        for _, trans_prefix in potential_prefix_combo:
+        for _, trans_prefix in POTENTIAL_PREFIX_COMBOS:
             translations = re.findall(f'{trans_prefix}\n*(.*?)(?:#\d+|<summary>|\n*$)', content, re.DOTALL)
             if translations:
                 return self._clean_translations(translations, content)
@@ -102,6 +102,7 @@ def translate_chunk(self, chunk_id: int, chunk: List[Tuple[int, str]],
 class ContextReviewerAgent(Agent):
     """
     Review the context of the subtitles to ensure accuracy and completeness.
+    TODO: Add chunking support.
     """
 
     TEMPERATURE = 0.8
@@ -149,7 +150,7 @@ def __init__(self, src_lang, target_lang, info: TranslateInfo = TranslateInfo(),
 
     def _parse_responses(self, resp) -> List[str]:
         content = self.chatbot.get_content(resp)
-        revised = re.findall(proofread_prefix + r'\s*(.*)', content, re.MULTILINE)
+        revised = re.findall(PROOFREAD_PREFIX + r'\s*(.*)', content, re.MULTILINE)
 
         return revised
 

diff --git a/openlrc/chatbot.py b/openlrc/chatbot.py
@@ -233,8 +233,8 @@ async def _create_achat(self, messages: List[Dict], output_checker: Callable = l
             except openai.APIConnectionError:
                 logger.warning(f'API connection error. Wait 15s before retry. Retry num: {i + 1}.')
                 time.sleep(15)
-            except openai.APIError:
-                logger.warning(f'API error. Wait 15s before retry. Retry num: {i + 1}.')
+            except openai.APIError as e:
+                logger.warning(f'API error: {e}. Wait 15s before retry. Retry num: {i + 1}.')
                 time.sleep(15)
 
         if not response: