init commit

This commit is contained in:
2026-03-11 07:21:21 -05:00
commit 5aff2183d6
29 changed files with 2270 additions and 0 deletions

0
app/services/__init__.py Normal file
View File

124
app/services/llm.py Normal file
View File

@@ -0,0 +1,124 @@
"""LLM client wrapper for Ollama / OpenAI-compatible endpoints."""
import json
from collections.abc import AsyncIterator
from contextlib import asynccontextmanager
from typing import Self
import httpx
from app.models.config import LLMConfig
from app.models.message import Message
from app.utils.logging import get_logger
logger = get_logger(__name__)
# --- Exception hierarchy ---
class LLMError(Exception):
"""Base exception for LLM client errors."""
class LLMConnectionError(LLMError):
"""Connection or timeout failure when reaching the LLM endpoint."""
class LLMResponseError(LLMError):
"""Non-2xx HTTP response from the LLM endpoint."""
def __init__(self, message: str, status_code: int | None = None) -> None:
super().__init__(message)
self.status_code = status_code
class LLMStreamError(LLMError):
"""Malformed SSE data during streaming."""
# --- Client ---
class LLMClient:
"""Async streaming client for OpenAI-compatible chat completions.
Designed for Ollama but works with any endpoint implementing the
OpenAI /v1/chat/completions SSE streaming protocol.
"""
def __init__(self, config: LLMConfig) -> None:
"""Initialize the LLM client.
Args:
config: LLM configuration (model, endpoint, timeout, etc.).
"""
self._config = config
self._client = httpx.AsyncClient(
base_url=config.endpoint,
timeout=httpx.Timeout(config.timeout, connect=10.0),
)
async def stream_chat(self, messages: list[Message]) -> AsyncIterator[dict]:
"""Stream a chat completion request, yielding parsed SSE chunks.
Args:
messages: Conversation history to send to the model.
Yields:
Parsed JSON dicts from each SSE data line.
Raises:
LLMConnectionError: On connection or timeout failures.
LLMResponseError: On non-2xx HTTP status.
LLMStreamError: On malformed SSE data (only if every line fails).
"""
payload = {
"model": self._config.model,
"messages": [m.to_api_dict() for m in messages],
"stream": True,
"temperature": self._config.temperature,
"max_tokens": self._config.max_tokens,
}
try:
async with self._client.stream(
"POST", self._config.api_path, json=payload
) as response:
if response.status_code != 200:
body = await response.aread()
raise LLMResponseError(
f"LLM returned {response.status_code}: {body.decode(errors='replace')}",
status_code=response.status_code,
)
async for line in response.aiter_lines():
if not line.startswith("data: "):
continue
data = line[6:] # strip "data: " prefix
if data.strip() == "[DONE]":
return
try:
yield json.loads(data)
except json.JSONDecodeError:
logger.warning("malformed_sse_chunk", data=data[:200])
except httpx.ConnectError as e:
raise LLMConnectionError(f"Cannot connect to LLM endpoint: {e}") from e
except httpx.TimeoutException as e:
raise LLMConnectionError(f"LLM request timed out: {e}") from e
except httpx.HTTPError as e:
raise LLMError(f"HTTP error communicating with LLM: {e}") from e
async def close(self) -> None:
"""Close the underlying HTTP client."""
await self._client.aclose()
async def __aenter__(self) -> Self:
return self
async def __aexit__(self, *exc: object) -> None:
await self.close()

150
app/services/streaming.py Normal file
View File

@@ -0,0 +1,150 @@
"""Streaming response handler — accumulates SSE chunks into a complete Message."""
from collections.abc import AsyncIterator
from rich.live import Live
from rich.markdown import Markdown
from app.models.config import DisplayConfig
from app.models.message import Message
from app.models.tool_call import ToolCall, ToolCallFunction
from app.utils.display import print_assistant_message
from app.utils.logging import console, get_logger
from app.utils.token_counter import TokenUsage
logger = get_logger(__name__)
class StreamHandler:
"""Processes an SSE chunk stream into a Rich live display and final Message.
Accumulates content deltas and tool call fragments, renders a live Markdown
panel during streaming, and produces a complete assistant Message on finish.
"""
def __init__(self, display_config: DisplayConfig) -> None:
"""Initialize the stream handler.
Args:
display_config: Display preferences (streaming toggle, etc.).
"""
self._display_config = display_config
self._accumulated_content: str = ""
self._accumulated_reasoning: str = ""
self._tool_calls: dict[int, dict[str, str]] = {}
self._usage: TokenUsage | None = None
async def process_stream(self, chunk_iter: AsyncIterator[dict]) -> Message:
"""Consume a chunk iterator, rendering live output and returning the final Message.
Args:
chunk_iter: Async iterator of parsed SSE chunk dicts.
Returns:
Complete assistant Message with accumulated content and tool calls.
"""
with Live(console=console, refresh_per_second=8) as live:
async for chunk in chunk_iter:
self._process_chunk(chunk)
# Show reasoning while waiting for content
display_text = self._accumulated_content
if not display_text and self._accumulated_reasoning:
display_text = f"*thinking...*"
if display_text and self._display_config.stream_output:
live.update(Markdown(display_text))
# Final static render
if self._accumulated_content:
print_assistant_message(self._accumulated_content)
tool_calls = self._build_tool_calls() or None
return Message(
role="assistant",
content=self._accumulated_content or None,
tool_calls=tool_calls,
)
def _process_chunk(self, chunk: dict) -> None:
"""Extract content, tool calls, and usage from a single SSE chunk.
Args:
chunk: Parsed JSON dict from one SSE data line.
"""
# Content delta
choices = chunk.get("choices", [])
if choices:
delta = choices[0].get("delta", {})
content_piece = delta.get("content")
if content_piece:
self._accumulated_content += content_piece
# Reasoning tokens (e.g. qwen3.5 thinking mode)
reasoning_piece = delta.get("reasoning")
if reasoning_piece:
self._accumulated_reasoning += reasoning_piece
# Tool call deltas (accumulated by index)
for tc_delta in delta.get("tool_calls", []):
idx = tc_delta.get("index", 0)
if idx not in self._tool_calls:
self._tool_calls[idx] = {
"id": tc_delta.get("id", ""),
"name": "",
"arguments": "",
}
entry = self._tool_calls[idx]
if tc_delta.get("id"):
entry["id"] = tc_delta["id"]
func = tc_delta.get("function", {})
if func.get("name"):
entry["name"] += func["name"]
if func.get("arguments"):
entry["arguments"] += func["arguments"]
# Token usage (typically in the final chunk)
usage_data = chunk.get("usage")
if usage_data:
self._usage = TokenUsage(
prompt_tokens=usage_data.get("prompt_tokens", 0),
completion_tokens=usage_data.get("completion_tokens", 0),
total_tokens=usage_data.get("total_tokens", 0),
)
def _build_tool_calls(self) -> list[ToolCall]:
"""Convert accumulated tool call fragments into sorted ToolCall list.
Returns:
List of ToolCall objects sorted by stream index.
"""
if not self._tool_calls:
return []
result: list[ToolCall] = []
for idx in sorted(self._tool_calls):
entry = self._tool_calls[idx]
result.append(
ToolCall(
id=entry["id"],
type="function",
function=ToolCallFunction(
name=entry["name"],
arguments=entry["arguments"],
),
)
)
return result
@property
def usage(self) -> TokenUsage | None:
"""Token usage reported by the API, if available."""
return self._usage
def reset(self) -> None:
"""Clear all accumulators for the next turn."""
self._accumulated_content = ""
self._accumulated_reasoning = ""
self._tool_calls.clear()
self._usage = None