core/agent+cli: treat last-request prompt_tokens as context size

context_tokens prefers API usage from the last model call; char est.
is only a pre-call fallback. Clarify billed turn/session totals in UI.
This commit is contained in:
2026-07-15 11:07:52 +08:00
parent a61b536243
commit c9e805f38d
6 changed files with 73 additions and 10 deletions
+15
View File
@@ -273,6 +273,21 @@ async def test_chat_agent_turn_usage_sums_tool_rounds() -> None:
assert agent.last_request_usage.prompt_tokens == 200
assert agent.last_turn_usage.prompt_tokens == 300
assert agent.last_turn_usage.total_tokens == 308
# Context size is last request prompt, not billed sum
assert agent.context_tokens == 200
assert agent.context_tokens_source == "api"
async def test_context_tokens_falls_back_to_message_estimate() -> None:
agent = ChatAgent(
ScriptedClient([]),
model="m",
stream=False,
messages=[UserChatMessage(content="12345678")],
)
assert agent.last_request_usage.is_zero()
assert agent.context_tokens_source == "estimate"
assert agent.context_tokens >= 1
async def test_stream_yields_deltas_incrementally() -> None:
+14 -1
View File
@@ -145,12 +145,25 @@ async def test_rounds(state: ReplState) -> None:
async def test_status_shows_context_tokens(
state: ReplState, capsys: pytest.CaptureFixture[str]
) -> None:
from plyngent.agent.usage import TokenUsage
from plyngent.lmproto.openai_compatible.model import UserChatMessage
state.agent.messages = [UserChatMessage(content="hello")]
assert await handle_slash(state, "/status") is True
out = capsys.readouterr().out
assert "context_tokens~=" in out
assert "context_tokens=" in out
assert "(est)" in out # no API usage yet
assert "context_chars=" in out
assert "tool_result_max=" in out
assert str(state.workspace) in out
state.agent.last_request_usage = TokenUsage(
prompt_tokens=1234,
completion_tokens=10,
total_tokens=1244,
source="api",
)
assert await handle_slash(state, "/status") is True
out2 = capsys.readouterr().out
assert "context_tokens=1234/" in out2
assert "(api)" in out2