test_llm.py

from click.testing import CliRunner
import datetime
import llm
from llm.cli import cli
from llm.migrations import migrate
import json
import os
import pytest
import re
import sqlite_utils
import sys
from ulid import ULID
from unittest import mock


def test_version():
    runner = CliRunner()
    with runner.isolated_filesystem():
        result = runner.invoke(cli, ["--version"])
        assert result.exit_code == 0
        assert result.output.startswith("cli, version ")


@pytest.fixture
def log_path(user_path):
    log_path = str(user_path / "logs.db")
    db = sqlite_utils.Database(log_path)
    migrate(db)
    start = datetime.datetime.utcnow()
    db["responses"].insert_all(
        {
            "id": str(ULID()).lower(),
            "system": "system",
            "prompt": "prompt",
            "response": "response",
            "model": "davinci",
            "datetime_utc": (start + datetime.timedelta(seconds=i)).isoformat(),
            "conversation_id": "abc123",
        }
        for i in range(100)
    )
    return log_path


datetime_re = re.compile(r"\d{4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2}")


def test_logs_text(log_path):
    runner = CliRunner()
    args = ["logs", "-p", str(log_path)]
    result = runner.invoke(cli, args, catch_exceptions=False)
    assert result.exit_code == 0
    output = result.output
    # Replace 2023-08-17T20:53:58 with YYYY-MM-DDTHH:MM:SS
    output = datetime_re.sub("YYYY-MM-DDTHH:MM:SS", output)

    assert output == (
        "# YYYY-MM-DDTHH:MM:SS    conversation: abc123\n\n"
        "Model: **davinci**\n\n"
        "## Prompt:\n\n"
        "prompt\n\n"
        "## System:\n\n"
        "system\n\n"
        "## Response:\n\n"
        "response\n\n"
        "# YYYY-MM-DDTHH:MM:SS    conversation: abc123\n\n"
        "Model: **davinci**\n\n"
        "## Prompt:\n\n"
        "prompt\n\n"
        "## Response:\n\n"
        "response\n\n"
        "# YYYY-MM-DDTHH:MM:SS    conversation: abc123\n\n"
        "Model: **davinci**\n\n"
        "## Prompt:\n\n"
        "prompt\n\n"
        "## Response:\n\n"
        "response\n\n"
    )


@pytest.mark.parametrize("n", (None, 0, 2))
def test_logs_json(n, log_path):
    "Test that logs command correctly returns requested -n records"
    runner = CliRunner()
    args = ["logs", "-p", str(log_path), "--json"]
    if n is not None:
        args.extend(["-n", str(n)])
    result = runner.invoke(cli, args, catch_exceptions=False)
    assert result.exit_code == 0
    logs = json.loads(result.output)
    expected_length = 3
    if n is not None:
        if n == 0:
            expected_length = 100
        else:
            expected_length = n
    assert len(logs) == expected_length


@pytest.mark.parametrize(
    "args", (["-r"], ["--response"], ["list", "-r"], ["list", "--response"])
)
def test_logs_response_only(args, log_path):
    "Test that logs -r/--response returns just the last response"
    runner = CliRunner()
    result = runner.invoke(cli, ["logs"] + args, catch_exceptions=False)
    assert result.exit_code == 0
    assert result.output == "response\n"


@pytest.mark.xfail(sys.platform == "win32", reason="Expected to fail on Windows")
@pytest.mark.parametrize("env", ({}, {"LLM_USER_PATH": "/tmp/llm-user-path"}))
def test_logs_path(monkeypatch, env, user_path):
    for key, value in env.items():
        monkeypatch.setenv(key, value)
    runner = CliRunner()
    result = runner.invoke(cli, ["logs", "path"])
    assert result.exit_code == 0
    if env:
        expected = env["LLM_USER_PATH"] + "/logs.db"
    else:
        expected = str(user_path) + "/logs.db"
    assert result.output.strip() == expected


@pytest.mark.parametrize("model", ("davinci", "curie"))
def test_logs_filtered(user_path, model):
    log_path = str(user_path / "logs.db")
    db = sqlite_utils.Database(log_path)
    migrate(db)
    db["responses"].insert_all(
        {
            "id": str(ULID()).lower(),
            "system": "system",
            "prompt": "prompt",
            "response": "response",
            "model": "davinci" if i % 2 == 0 else "curie",
        }
        for i in range(100)
    )
    runner = CliRunner()
    result = runner.invoke(cli, ["logs", "list", "-m", model, "--json"])
    assert result.exit_code == 0
    records = json.loads(result.output.strip())
    assert all(record["model"] == model for record in records)


@pytest.mark.parametrize(
    "query,expected",
    (
        # With no search term order should be by datetime
        ("", ["doc1", "doc2", "doc3"]),
        # With a search it's order by rank instead
        ("llama", ["doc1", "doc3"]),
        ("alpaca", ["doc2"]),
    ),
)
def test_logs_search(user_path, query, expected):
    log_path = str(user_path / "logs.db")
    db = sqlite_utils.Database(log_path)
    migrate(db)

    def _insert(id, text):
        db["responses"].insert(
            {
                "id": id,
                "system": "system",
                "prompt": text,
                "response": "response",
                "model": "davinci",
            }
        )

    _insert("doc1", "llama")
    _insert("doc2", "alpaca")
    _insert("doc3", "llama llama")
    runner = CliRunner()
    result = runner.invoke(cli, ["logs", "list", "-q", query, "--json"])
    assert result.exit_code == 0
    records = json.loads(result.output.strip())
    assert [record["id"] for record in records] == expected


def test_llm_prompt_creates_log_database(mocked_openai_chat, tmpdir, monkeypatch):
    user_path = tmpdir / "user"
    monkeypatch.setenv("LLM_USER_PATH", str(user_path))
    runner = CliRunner()
    result = runner.invoke(
        cli,
        ["three names \nfor a pet pelican", "--no-stream", "--key", "x"],
        catch_exceptions=False,
    )
    assert result.exit_code == 0
    assert result.output == "Bob, Alice, Eve\n"
    # Should have created user_path and put a logs.db in it
    assert (user_path / "logs.db").exists()
    assert sqlite_utils.Database(str(user_path / "logs.db"))["responses"].count == 1


@mock.patch.dict(os.environ, {"OPENAI_API_KEY": "X"})
@pytest.mark.parametrize("use_stdin", (True, False, "split"))
@pytest.mark.parametrize(
    "logs_off,logs_args,should_log",
    (
        (True, [], False),
        (False, [], True),
        (False, ["--no-log"], False),
        (False, ["--log"], True),
        (True, ["-n"], False),  # Short for --no-log
        (True, ["--log"], True),
    ),
)
def test_llm_default_prompt(
    mocked_openai_chat, use_stdin, user_path, logs_off, logs_args, should_log
):
    # Reset the log_path database
    log_path = user_path / "logs.db"
    log_db = sqlite_utils.Database(str(log_path))
    log_db["responses"].delete_where()

    logs_off_path = user_path / "logs-off"
    if logs_off:
        # Turn off logging
        assert not logs_off_path.exists()
        CliRunner().invoke(cli, ["logs", "off"])
        assert logs_off_path.exists()
    else:
        # Turn on logging
        CliRunner().invoke(cli, ["logs", "on"])
        assert not logs_off_path.exists()

    # Run the prompt
    runner = CliRunner()
    prompt = "three names \nfor a pet pelican"
    input = None
    args = ["--no-stream"]
    if use_stdin == "split":
        input = "three names"
        args.append("\nfor a pet pelican")
    elif use_stdin:
        input = prompt
    else:
        args.append(prompt)
    args += logs_args
    result = runner.invoke(cli, args, input=input, catch_exceptions=False)
    assert result.exit_code == 0
    assert result.output == "Bob, Alice, Eve\n"
    last_request = mocked_openai_chat.get_requests()[-1]
    assert last_request.headers["Authorization"] == "Bearer X"

    # Was it logged?
    rows = list(log_db["responses"].rows)

    if not should_log:
        assert len(rows) == 0
        return

    assert len(rows) == 1
    expected = {
        "model": "gpt-3.5-turbo",
        "prompt": "three names \nfor a pet pelican",
        "system": None,
        "options_json": "{}",
        "response": "Bob, Alice, Eve",
    }
    row = rows[0]
    assert expected.items() <= row.items()
    assert isinstance(row["duration_ms"], int)
    assert isinstance(row["datetime_utc"], str)
    assert json.loads(row["prompt_json"]) == {
        "messages": [{"role": "user", "content": "three names \nfor a pet pelican"}]
    }
    assert json.loads(row["response_json"]) == {
        "model": "gpt-3.5-turbo",
        "choices": [{"message": {"content": "Bob, Alice, Eve"}}],
    }

    # Test "llm logs"
    log_result = runner.invoke(
        cli, ["logs", "-n", "1", "--json"], catch_exceptions=False
    )
    log_json = json.loads(log_result.output)

    # Should have logged correctly:
    assert (
        log_json[0].items()
        >= {
            "model": "gpt-3.5-turbo",
            "prompt": "three names \nfor a pet pelican",
            "system": None,
            "prompt_json": {
                "messages": [
                    {"role": "user", "content": "three names \nfor a pet pelican"}
                ]
            },
            "options_json": {},
            "response": "Bob, Alice, Eve",
            "response_json": {
                "model": "gpt-3.5-turbo",
                "choices": [{"message": {"content": "Bob, Alice, Eve"}}],
            },
            # This doesn't have the \n after three names:
            "conversation_name": "three names for a pet pelican",
            "conversation_model": "gpt-3.5-turbo",
        }.items()
    )


def test_openai_chat_stream(mocked_openai_chat_stream, user_path):
    runner = CliRunner()
    result = runner.invoke(cli, ["-m", "gpt-3.5-turbo", "--key", "x", "Say hi"])
    assert result.exit_code == 0
    assert result.output == "Hi.\n"


def test_openai_completion(mocked_openai_completion, user_path):
    log_path = user_path / "logs.db"
    log_db = sqlite_utils.Database(str(log_path))
    log_db["responses"].delete_where()
    runner = CliRunner()
    result = runner.invoke(
        cli,
        [
            "-m",
            "gpt-3.5-turbo-instruct",
            "Say this is a test",
            "--no-stream",
            "--key",
            "x",
        ],
        catch_exceptions=False,
    )
    assert result.exit_code == 0
    assert result.output == "\n\nThis is indeed a test\n"

    # Should have requested 256 tokens
    last_request = mocked_openai_completion.get_requests()[-1]
    assert json.loads(last_request.content) == {
        "model": "gpt-3.5-turbo-instruct",
        "prompt": "Say this is a test",
        "stream": False,
        "max_tokens": 256,
    }

    # Check it was logged
    rows = list(log_db["responses"].rows)
    assert len(rows) == 1
    expected = {
        "model": "gpt-3.5-turbo-instruct",
        "prompt": "Say this is a test",
        "system": None,
        "prompt_json": '{"messages": ["Say this is a test"]}',
        "options_json": "{}",
        "response": "\n\nThis is indeed a test",
    }
    row = rows[0]
    assert expected.items() <= row.items()


def test_openai_completion_system_prompt_error():
    runner = CliRunner()
    result = runner.invoke(
        cli,
        [
            "-m",
            "gpt-3.5-turbo-instruct",
            "Say this is a test",
            "--no-stream",
            "--key",
            "x",
            "--system",
            "system prompts not allowed",
        ],
        catch_exceptions=False,
    )
    assert result.exit_code == 1
    assert (
        result.output
        == "Error: System prompts are not supported for OpenAI completion models\n"
    )


def test_openai_completion_logprobs_stream(
    mocked_openai_completion_logprobs_stream, user_path
):
    log_path = user_path / "logs.db"
    log_db = sqlite_utils.Database(str(log_path))
    log_db["responses"].delete_where()
    runner = CliRunner()
    args = [
        "-m",
        "gpt-3.5-turbo-instruct",
        "Say hi",
        "-o",
        "logprobs",
        "2",
        "--key",
        "x",
    ]
    result = runner.invoke(cli, args, catch_exceptions=False)
    assert result.exit_code == 0
    assert result.output == "\n\nHi.\n"
    rows = list(log_db["responses"].rows)
    assert len(rows) == 1
    row = rows[0]
    assert json.loads(row["response_json"]) == {
        "content": "\n\nHi.",
        "logprobs": [
            {"text": "\n\n", "top_logprobs": [{"\n\n": -0.6, "\n": -1.9}]},
            {"text": "Hi", "top_logprobs": [{"Hi": -1.1, "Hello": -0.7}]},
            {"text": ".", "top_logprobs": [{".": -1.1, "!": -0.9}]},
            {"text": "", "top_logprobs": []},
        ],
        "id": "cmpl-80MdSaou7NnPuff5ZyRMysWBmgSPS",
        "object": "text_completion",
        "model": "gpt-3.5-turbo-instruct",
        "created": 1695097702,
    }


def test_openai_completion_logprobs_nostream(
    mocked_openai_completion_logprobs, user_path
):
    log_path = user_path / "logs.db"
    log_db = sqlite_utils.Database(str(log_path))
    log_db["responses"].delete_where()
    runner = CliRunner()
    args = [
        "-m",
        "gpt-3.5-turbo-instruct",
        "Say hi",
        "-o",
        "logprobs",
        "2",
        "--key",
        "x",
        "--no-stream",
    ]
    result = runner.invoke(cli, args, catch_exceptions=False)
    assert result.exit_code == 0
    assert result.output == "\n\nHi.\n"
    rows = list(log_db["responses"].rows)
    assert len(rows) == 1
    row = rows[0]
    assert json.loads(row["response_json"]) == {
        "choices": [
            {
                "finish_reason": "stop",
                "index": 0,
                "logprobs": {
                    "text_offset": [16, 18, 20],
                    "token_logprobs": [-0.6, -1.1, -0.9],
                    "tokens": ["\n\n", "Hi", "1"],
                    "top_logprobs": [
                        {"\n": -1.9, "\n\n": -0.6},
                        {"Hello": -0.7, "Hi": -1.1},
                        {"!": -1.1, ".": -0.9},
                    ],
                },
                "text": "\n\nHi.",
            }
        ],
        "created": 1695097747,
        "id": "cmpl-80MeBfKJutM0uMNJkRrebJLeP3bxL",
        "model": "gpt-3.5-turbo-instruct",
        "object": "text_completion",
        "usage": {"completion_tokens": 3, "prompt_tokens": 5, "total_tokens": 8},
    }


EXTRA_MODELS_YAML = """
- model_id: orca
  model_name: orca-mini-3b
  api_base: "http://localai.localhost"
- model_id: completion-babbage
  model_name: babbage
  api_base: "http://localai.localhost"
  completion: 1
"""


def test_openai_localai_configuration(mocked_localai, user_path):
    log_path = user_path / "logs.db"
    sqlite_utils.Database(str(log_path))
    # Write the configuration file
    config_path = user_path / "extra-openai-models.yaml"
    config_path.write_text(EXTRA_MODELS_YAML, "utf-8")
    # Run the prompt
    runner = CliRunner()
    prompt = "three names \nfor a pet pelican"
    result = runner.invoke(cli, ["--no-stream", "--model", "orca", prompt])
    assert result.exit_code == 0
    assert result.output == "Bob, Alice, Eve\n"
    last_request = mocked_localai.get_requests()[-1]
    assert json.loads(last_request.content) == {
        "model": "orca-mini-3b",
        "messages": [{"role": "user", "content": "three names \nfor a pet pelican"}],
        "stream": False,
    }
    # And check the completion model too
    result2 = runner.invoke(cli, ["--no-stream", "--model", "completion-babbage", "hi"])
    assert result2.exit_code == 0
    assert result2.output == "Hello\n"
    last_request2 = mocked_localai.get_requests()[-1]
    assert json.loads(last_request2.content) == {
        "model": "babbage",
        "prompt": "hi",
        "stream": False,
    }


EXPECTED_OPTIONS = """
OpenAI Chat: gpt-3.5-turbo (aliases: 3.5, chatgpt)
  temperature: float
    What sampling temperature to use, between 0 and 2. Higher values like
    0.8 will make the output more random, while lower values like 0.2 will
    make it more focused and deterministic.
  max_tokens: int
    Maximum number of tokens to generate.
  top_p: float
    An alternative to sampling with temperature, called nucleus sampling,
    where the model considers the results of the tokens with top_p
    probability mass. So 0.1 means only the tokens comprising the top 10%
    probability mass are considered. Recommended to use top_p or
    temperature but not both.
  frequency_penalty: float
    Number between -2.0 and 2.0. Positive values penalize new tokens based
    on their existing frequency in the text so far, decreasing the model's
    likelihood to repeat the same line verbatim.
  presence_penalty: float
    Number between -2.0 and 2.0. Positive values penalize new tokens based
    on whether they appear in the text so far, increasing the model's
    likelihood to talk about new topics.
  stop: str
    A string where the API will stop generating further tokens.
  logit_bias: dict, str
    Modify the likelihood of specified tokens appearing in the completion.
    Pass a JSON string like '{"1712":-100, "892":-100, "1489":-100}'
"""


def test_llm_models_options(user_path):
    runner = CliRunner()
    result = runner.invoke(cli, ["models", "--options"], catch_exceptions=False)
    assert result.exit_code == 0
    assert EXPECTED_OPTIONS.strip() in result.output


def test_llm_user_dir(tmpdir, monkeypatch):
    user_dir = str(tmpdir / "u")
    monkeypatch.setenv("LLM_USER_PATH", user_dir)
    assert not os.path.exists(user_dir)
    user_dir2 = llm.user_dir()
    assert user_dir == str(user_dir2)
    assert os.path.exists(user_dir)