test: improved testing for utils, removed unused functions

gptme/cli.py

@@ -51,7 +51,7 @@
 from .prompts import get_prompt
 from .tabcomplete import register_tabcomplete
 from .tools import execute_msg, init_tools
-from .util import epoch_to_age, generate_unique_name
+from .util import epoch_to_age, generate_name
 
 logger = logging.getLogger(__name__)
 print_builtin = __builtins__["print"]  # type: ignore

gptme/util.py

@@ -36,20 +36,6 @@ def get_tokenizer(model: str):
     return tiktoken.get_encoding("cl100k_base")
 
 
-def len_tokens_approx(content: str | list[Message]) -> int:
-    """Approximate the number of tokens in a string by assuming tokens have len 3 (lol)."""
-    if isinstance(content, list):
-        return sum(len_tokens_approx(msg.content) for msg in content)
-    return len(content) // 3
-
-
-def msgs2text(msgs: list[Message]) -> str:
-    output = ""
-    for msg in msgs:
-        output += f"{msg.role.capitalize()}: {msg.content}\n\n"
-    return output
-
-
 def msgs2dicts(msgs: list[Message]) -> list[dict]:
     """Convert a list of Message objects to a list of dicts ready to pass to an LLM."""
     return [msg.to_dict(keys=["role", "content"]) for msg in msgs]
@@ -66,10 +52,13 @@ def msgs2dicts(msgs: list[Message]) -> list[dict]:
     "crawling",
     "sneaking",
     "sprinting",
+    "sneaking",
+    "dancing",
+    "singing",
+    "laughing",
 ]
 adjectives = [
     "funny",
-    "red",
     "happy",
     "sad",
     "angry",
@@ -78,34 +67,55 @@ def msgs2dicts(msgs: list[Message]) -> list[dict]:
     "sneaky",
     "sleepy",
     "hungry",
+    # colors
+    "red",
+    "blue",
+    "green",
+    "pink",
+    "purple",
+    "yellow",
+    "orange",
 ]
 nouns = [
-    "walrus",
-    "pelican",
     "cat",
     "dog",
-    "elephant",
     "rat",
     "mouse",
-    "bird",
     "fish",
+    "elephant",
+    "dinosaur",
+    # birds
+    "bird",
+    "pelican",
+    # fictional
     "dragon",
     "unicorn",
-    "dinosaur",
+    "mermaid",
+    "monster",
+    "alien",
+    "robot",
+    # sea creatures
+    "whale",
+    "shark",
+    "walrus",
+    "octopus",
+    "squid",
+    "jellyfish",
+    "starfish",
+    "penguin",
+    "seal",
 ]
 
 
-def generate_unique_name():
+def generate_name():
     action = random.choice(actions)
     adjective = random.choice(adjectives)
     noun = random.choice(nouns)
-    unique_name = f"{action}-{adjective}-{noun}"
-    assert is_generated_name(unique_name)
-    return unique_name
+    return f"{action}-{adjective}-{noun}"
 
 
 def is_generated_name(name: str) -> bool:
-    """if name is a name generated by generate_unique_name"""
+    """if name is a name generated by generate_name"""
     all_words = actions + adjectives + nouns
     return name.count("-") == 2 and all(word in all_words for word in name.split("-"))
 
@@ -125,14 +135,14 @@ def epoch_to_age(epoch):
         return f"{age.days} days ago ({datetime.fromtimestamp(epoch).strftime('%Y-%m-%d')})"
 
 
-def print_preview(code: str, lang: str):
+def print_preview(code: str, lang: str):  # pragma: no cover
     print()
     print("[bold white]Preview[/bold white]")
     print(Syntax(code.strip(), lang))
     print()
 
 
-def ask_execute(question="Execute code?", default=True) -> bool:
+def ask_execute(question="Execute code?", default=True) -> bool:  # pragma: no cover
     # TODO: add a way to outsource ask_execute decision to another agent/LLM
     console = Console()
     choicestr = f"({'Y' if default else 'y'}/{'n' if default else 'N'})"

tests/test_util.py

@@ -1,4 +1,19 @@
+from datetime import datetime
+
 from gptme.tools import is_supported_codeblock
+from gptme.util import epoch_to_age, generate_name, is_generated_name
+
+
+def test_generate_name():
+    name = generate_name()
+    assert is_generated_name(name)
+
+
+def test_epoch_to_age():
+    epoch_today = datetime.now().timestamp()
+    assert epoch_to_age(epoch_today) == "just now"
+    epoch_yesterday = epoch_today - 24 * 60 * 60
+    assert epoch_to_age(epoch_yesterday) == "yesterday"
 
 
 def test_is_supported_codeblock():