From 043b6128c276f19769e24587ed7258951a58e848 Mon Sep 17 00:00:00 2001
From: jstzwj <1103870790@qq.com>
Date: Wed, 27 Dec 2023 03:40:33 +0800
Subject: [PATCH] update readme

---
 README.md             | 25 +++++++++++++++++++++----
 chatproto/registry.py |  2 ++
 2 files changed, 23 insertions(+), 4 deletions(-)

diff --git a/README.md b/README.md
index 6be580c..1596b98 100644
--- a/README.md
+++ b/README.md
@@ -3,21 +3,38 @@ Large Language Model Chat Protocol.
 
 The different chat prompt formats used by different Large Language Models have been a problem for developers. We developed `chatproto` to output the prompt format for different LLMs through a unified interface.
 
+Compared to the `apply_chat_format` function in HuggingFace and the version in FastChat, `ChatProto` can locate the position of each message after applying the template. This makes it very convenient for us to mask out certain conversations during training.
+
 ## Quick Start
 ```python
 from chatproto.conversation.history import ConversationHistory
-from chatproto.conversation.models.baichuan import baichuan
+from chatproto.registry import list_conv_settings, get_conv_settings
+
+# Print all available settings
+all_settings = list_conv_settings()
+print(all_settings)
 
+settings = get_conv_settings("openbuddy")
 history = ConversationHistory(
     "SYSTEM_MESSAGE",
     messages=[
-        (baichuan.roles[0], "user"),
-        (baichuan.roles[1], "assistant"),
+        (settings.roles[0], "Hello!"),
+        (settings.roles[1], "Hello! How can I assist you today?"),
     ],
     offset=0,
-    settings=baichuan
+    settings=settings
 )
+# Apply the template
 print(history.get_prompt())
+
+# Get prompt and indices
+prompt, indices = history.get_prompt_and_indices()
+# Print the start and end offsets of each message in the conversation one by one.
+# The start and end offsets here refer to the offsets in the text, not the tokens.
+# They do not include any additional characters added in the template.
+system_start, system_end = indices[0]
+for i, (conv_start, conv_end) in enumerate(indices[1:]):
+    print((conv_start, conv_end))
 ```
 
 ## Install
diff --git a/chatproto/registry.py b/chatproto/registry.py
index e1bf3ef..aa1e97b 100644
--- a/chatproto/registry.py
+++ b/chatproto/registry.py
@@ -22,6 +22,8 @@ def get_conv_settings(name: str) -> ConversationSettings:
     """Get a ConversationSettings template."""
     return conv_settings[name].copy()
 
+def list_conv_settings() -> List[str]:
+    return list(conv_settings.keys())
 
 settings_path = os.path.join(os.path.dirname(__file__), "conversation", "models")
 for module_loader, name, ispkg in pkgutil.iter_modules([settings_path]):