Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.
Python agentaianthropicclaude-codecompressioncontext-engineeringcontext-windowcursorfastapilangchainllmmcpopenaiprompt-engineeringproxypythonragtoken-optimizationtokenstypescript
"""Tests for inline_extractor.inject_memory_instruction system-prompt handling.
The system prompt's ``content`` can be a plain string or a list of content-part
dicts (OpenAI allows a list; Anthropic system prompts are commonly a list of
``{"type": "text", ...}`` blocks). The instruction must be appended without
crashing on the list form.
"""
from __future__ import annotations
from headroom.memory.inline_extractor import (
MEMORY_INSTRUCTION,
MEMORY_INSTRUCTION_SHORT,
inject_memory_instruction,
)
def test_str_system_content_is_concatenated() -> None:
out = inject_memory_instruction(
[{"role": "system", "content": "You are X."}, {"role": "user", "content": "hi"}],
short=True,
)
assert out[0]["content"] == "You are X." + MEMORY_INSTRUCTION_SHORT
def test_list_system_content_appends_a_text_part() -> None:
"""Regression: a list-shaped system content used to raise
``TypeError: can only concatenate list (not "str") to list``."""
out = inject_memory_instruction(
[
{"role": "system", "content": [{"type": "text", "text": "You are X."}]},
{"role": "user", "content": "hi"},
],
short=True,
)
content = out[0]["content"]
assert isinstance(content, list)
assert content[0] == {"type": "text", "text": "You are X."}
assert content[-1] == {"type": "text", "text": MEMORY_INSTRUCTION_SHORT}
# The original list is not mutated in place.
assert len(content) == 2
def test_list_system_content_long_instruction() -> None:
out = inject_memory_instruction(
[{"role": "system", "content": [{"type": "text", "text": "sys"}]}],
short=False,
)
assert out[0]["content"][-1] == {"type": "text", "text": MEMORY_INSTRUCTION}
def test_missing_system_prepends_default() -> None:
out = inject_memory_instruction([{"role": "user", "content": "hi"}], short=True)
assert out[0]["role"] == "system"
assert MEMORY_INSTRUCTION_SHORT in out[0]["content"]
def test_original_messages_not_mutated() -> None:
original = [{"role": "system", "content": [{"type": "text", "text": "sys"}]}]
inject_memory_instruction(original, short=True)
assert original[0]["content"] == [{"type": "text", "text": "sys"}]