fix ctrl+y don't works

pyproject.toml

@@ -10,6 +10,7 @@ description = "Chat with LLM in your terminal, be it shell generator, story tell
 readme = "README.md"
 keywords = ["llm", "shell", "gpt"]
 license = "GPL-3.0"
+requires-python = ">=3.0.0"
 dependencies = [
   "requests",
   "pyperclip",

shgpt/api/ollama.py

@@ -1,39 +1,39 @@
 import json
-import os
 from ..utils.http import TimeoutSession
 from ..utils.common import *
 from ..utils.conf import *
-from .history import DummyHistory, FileHistory
-
-HIST_SEP = "=========="
 
 
 # https://github.com/ollama/ollama/blob/main/docs/api.md#generate-a-completion
 class Ollama(object):
     def __init__(self, base_url, model, role, timeout):
         self.base_url = base_url
         self.http_session = TimeoutSession(timeout=timeout)
-        if ENABLE_HISTORY:
-            self.history_file = FileHistory(os.path.join(CONF_PATH, "history"))
-        else:
-            self.history_file = DummyHistory()
         self.model = model
         self.role = role
+        self.system_message = (
+            None
+            if role == "default"
+            else {"role": "system", "content": ROLE_CONTENT.get(self.role)}
+        )
+        self.messages = []
 
-    def generate(self, prompt, stream=True):
+    def chat(self, prompt, stream=True):
         url = self.base_url + "/api/chat"
         debug_print(
             f"generate: {prompt} to {url} with model {self.model} role {self.role} and stream {stream}"
         )
-        system_content = ROLE_CONTENT.get(self.role, self.role)
+        self.messages.append({"role": "user", "content": prompt})
+        if len(self.messages) > 10:
+            self.messages = self.messages[-10:]
         payload = {
-            "messages": [
-                {"role": "system", "content": system_content, "name": "ShellGPT"},
-                {"role": "user", "content": prompt, "name": "user"},
-            ],
+            "messages": [] if self.system_message is None else [self.system_message],
             "model": self.model,
             "stream": stream,
         }
+        for m in self.messages:
+            payload["messages"].append(m)
+
         debug_print(f"Infer message: {payload}")
         r = self.http_session.post(url, json=payload, stream=stream)
         if r.status_code != 200:
@@ -46,10 +46,3 @@ def generate(self, prompt, stream=True):
                 content = resp["message"]["content"]
                 answer += content
                 yield content
-            else:
-                self.history_file.write(rf"""{now_ms()},{resp['eval_duration']},{resp['eval_count']}
-{prompt}
-{HIST_SEP}
-{answer}
-{HIST_SEP}
-""")

shgpt/app.py

@@ -1,6 +1,6 @@
 import argparse
 import sys
-from os import makedirs
+from os import makedirs, path
 from .api.ollama import Ollama
 from .version import __version__
 from .utils.conf import *
@@ -13,6 +13,21 @@ def init_app():
     makedirs(CONF_PATH, exist_ok=True)
 
 
+def setup_readline():
+    import readline
+    import atexit
+
+    history = path.join(CONF_PATH, "history")
+    try:
+        readline.read_history_file(history)
+        readline.set_history_length(MAX_HISTORY)
+    except FileNotFoundError:
+        debug_print(f"History file not found: {history}")
+        pass
+
+    atexit.register(readline.write_history_file, history)
+
+
 def read_action(cmd):
     if IS_TTY:
         action = input("(E)xecute, (Y)ank or Continue(default): ")
@@ -25,7 +40,6 @@ def read_action(cmd):
 
 class ShellGPT(object):
     def __init__(self, url, model, role, timeout):
-        self.role = role
         self.is_shell = role == "shell"
         self.llm = Ollama(url, model, role, timeout)
 
@@ -54,7 +68,7 @@ def infer(self, prompt):
 
         buf = ""
         try:
-            for r in self.llm.generate(prompt):
+            for r in self.llm.chat(prompt):
                 buf += r
                 if self.is_shell is False:
                     print(r, end="")
@@ -78,6 +92,7 @@ def main():
         prog="shgpt",
         description="Chat with LLM in your terminal, be it shell generator, story teller, linux-terminal, etc.",
     )
+
     parser.add_argument(
         "-V", "--version", action="version", version="%(prog)s " + __version__
     )
@@ -151,4 +166,5 @@ def main():
     elif app_mode == AppMode.TUI:
         sg.tui(prompt)
     else:
+        setup_readline()
         sg.repl(prompt)

shgpt/tui/app.py

@@ -1,4 +1,4 @@
-from textual.app import App, ComposeResult
+from textual.app import App, ComposeResult, Binding
 from textual.widgets import Header, Footer, Static, TextArea, Button
 from typing import Optional
 from ..utils.common import *
@@ -24,9 +24,9 @@ def compose(self) -> ComposeResult:
 
 
 class ButtonDispatch(Static):
-    def __init__(self, copy_handler, run_handler):
+    def __init__(self, yank_handler, run_handler):
         super().__init__()
-        self.copy_handler = copy_handler
+        self.yank_handler = yank_handler
         self.run_handler = run_handler
 
     def compose(self) -> ComposeResult:
@@ -36,18 +36,18 @@ def compose(self) -> ComposeResult:
     def on_button_pressed(self, event: Button.Pressed) -> None:
         """Event handler called when a button is pressed."""
         if event.button.id == "copy":
-            self.copy_handler()
+            self.yank_handler()
         elif event.button.id == "run":
             self.run_handler()
 
 
 class ShellGPTApp(App):
     CSS_PATH = "app.tcss"
     BINDINGS = [
-        ("ctrl+j", "infer", "Infer answer"),
-        ("ctrl+d", "toggle_dark", "Toggle dark mode"),
-        ("ctrl+y", "Yank", "Yank code block"),
-        ("ctrl+r", "run", "Run code block"),
+        Binding("ctrl+j", "infer", "Infer answer"),
+        Binding("ctrl+d", "toggle_dark", "Toggle dark mode"),
+        Binding("ctrl+y", "yank", "Yank code block", priority=True),
+        Binding("ctrl+r", "run", "Run code block"),
     ]
 
     def __init__(self, llm, initial_prompt):
@@ -62,7 +62,7 @@ def compose(self) -> ComposeResult:
         yield PromptInput(self.initial_prompt)
         yield AnswerOutput()
         yield ButtonDispatch(
-            lambda: self.action_copy(),
+            lambda: self.action_yank(),
             lambda: self.action_run(),
         )
         yield CommandOutput()
@@ -79,7 +79,7 @@ def action_infer(self) -> None:
 
         self.has_inflight_req = True
         try:
-            self.action_infer_inner()
+            self.infer_inner()
         except Exception as e:
             answer_output = self.query_one("#answer_output")
             answer_output.load_text(f"Error when infer: {e}")
@@ -96,14 +96,14 @@ def get_answer_output(self) -> Optional[str]:
         text = out.text.strip()
         return None if text == "" else text
 
-    def action_infer_inner(self) -> None:
+    def infer_inner(self) -> None:
         prompt = self.get_prompt_input()
         if prompt is None:
             return
 
         debug_print(f"infer {prompt}")
         # llm infer
-        resp = self.llm.generate(prompt, True)
+        resp = self.llm.chat(prompt, True)
         buf = ""
         for item in resp:
             buf += item
@@ -112,7 +112,7 @@ def action_infer_inner(self) -> None:
         answer_output = self.query_one("#answer_output")
         answer_output.load_text(buf)
 
-    def action_copy(self) -> None:
+    def action_yank(self) -> None:
         text = self.get_answer_output()
         if text is None:
             return

shgpt/utils/conf.py

@@ -7,11 +7,11 @@
 OLLAMA_URL = environ.get("SHELLGPT_OLLAMA_URL", "http://127.0.0.1:11434")
 OLLAMA_MODEL = environ.get("SHELLGPT_OLLAMA_MODEL", "llama3")
 INFER_TIMEOUT = int(environ.get("SHELLGPT_INFER_TIMEOUT", "15"))  # seconds
-ENABLE_HISTORY = int(environ.get("SHELLGPT_ENABLE_HISTORY", "0")) == 1
+MAX_HISTORY = int(environ.get("SHELLGPT_MAX_HISTORY", "1000"))
 
-# There are different roles for different types of prompts
+# Built-in roles for different workloads.
 ROLE_CONTENT = {
-    "default": "",
+    "default": None,
     "code": """
 Provide only code as output without any description.
 Provide only code in plain text format without Markdown formatting.