Files
nofx/agent/planner_tools_test.go
tinkle-community 1851508353 feat(agent): make the assistant agentic - visible tools, LLM voice, full toolset
The agent felt like an artificial idiot because the LLM almost never spoke
for itself: 14+ Go paths injected fmt.Sprintf canned replies, the frontend
filtered out tool-progress events so users saw three dots for 10-20s, the
main prompt told the LLM "be a trading partner" AND "answer only what's
asked", and the planner sliced the toolset by inferred domain so a "BTC
dropped, how much am I losing?" question couldn't see positions and market
at the same time.

- agent/central_brain.go: shouldTrustDeterministicSkillReply now always
  returns false. Successful mutations (trader/strategy/model/exchange
  create/update/start/stop/delete) flow through reviewTaskCompletion so the
  LLM sees the real outcome JSON and writes the user-facing prose. The
  trade-confirmation regex path (handleTradeConfirmation) was already
  outside this code path and is unaffected.

- agent/agent.go: rewrite the Behavior section of the main system prompt.
  Replace the contradictory "answer only what's asked / don't upsell" with
  "lead with the direct answer, then optionally one relevant follow-up
  only when (a) open risk, (b) missing config, or (c) the next step is
  obvious — e.g. created, want me to start it?". Explicitly authorize
  chaining ("if the user says create and start, do both this turn") and
  ban "please wait / I'll get back to you" language because there is no
  background job to come back from.

- agent/tools.go: plannerToolsForText always returns the full 22-tool set
  (new __all__ domain). The old per-domain trimming hid manage_trader from
  market questions and execute_trade from anything that didn't look like
  an explicit trade — cross-domain reasoning was structurally blocked. The
  compact-vs-full strategy schema switch is preserved so mutation intents
  still see the full config schema.

- web/src/components/agent/{AgentStepPanel,ChatMessages}.tsx: stop
  filtering tool: steps. Map raw tool names to friendly labels with emoji
  ("get_positions" → "📊 检查持仓") in zh/en/id. Users now see what the
  agent is doing in real time instead of silence. central_brain routing
  chatter still gets dropped.

- agent/planner_tools_test.go: tests updated to assert the new
  full-toolset behavior and the compact-vs-full strategy schema switch.
2026-05-29 22:13:05 +08:00

104 lines
3.4 KiB
Go

package agent
import (
"encoding/json"
"testing"
"nofx/mcp"
)
// plannerToolsForText now always returns the FULL toolset (no per-domain
// trimming) so the LLM can cross-domain reason. The old "if market intent,
// hide manage_trader" filter was making cross-domain questions like "BTC
// dropped, how much am I losing?" impossible to answer because the agent
// couldn't see both market AND position tools in the same turn.
//
// We still trim the giant strategy schema for non-mutation intents because
// that one is genuinely huge and uninformative for read-only use.
func TestPlannerToolsExposeFullSetForMarketIntent(t *testing.T) {
tools := plannerToolsForText("看一下 BTCUSDT 行情和 K线")
names := toolNamesForTest(tools)
// Market tools must be present.
for _, expected := range []string{"get_market_snapshot", "get_market_price", "get_kline"} {
if !containsString(names, expected) {
t.Fatalf("expected market tool %q in %v", expected, names)
}
}
// Cross-domain tools (positions, balance, trader management) must ALSO be
// present so the agent can answer "how much am I losing" follow-ups
// without losing the market context.
for _, expected := range []string{"get_positions", "get_balance", "manage_trader"} {
if !containsString(names, expected) {
t.Fatalf("expected cross-domain tool %q in market context %v", expected, names)
}
}
}
func TestPlannerToolsExposeFullSetForExchangeIntent(t *testing.T) {
tools := plannerToolsForText("帮我添加 okx 交易所 API key")
names := toolNamesForTest(tools)
// At least the exchange management tools must show up.
for _, expected := range []string{"get_exchange_configs", "manage_exchange_config"} {
if !containsString(names, expected) {
t.Fatalf("expected exchange tool %q in %v", expected, names)
}
}
// And the agent still has the broader surface available — adding an
// exchange often leads to "now create a trader" so trader/strategy tools
// must be reachable in the same turn.
for _, expected := range []string{"manage_trader", "get_strategies"} {
if !containsString(names, expected) {
t.Fatalf("expected adjacent tool %q in exchange context %v", expected, names)
}
}
}
func TestPlannerToolsUseCompactManageStrategyForReadIntent(t *testing.T) {
tools := plannerToolsForText("列出我的策略")
tool := findToolForTest(tools, "manage_strategy")
if tool == nil {
t.Fatalf("expected manage_strategy in strategy tools")
}
raw, _ := json.Marshal(tool.Function.Parameters)
if len(raw) > 900 {
t.Fatalf("expected compact strategy schema, got %d bytes", len(raw))
}
if string(raw) == "" || !json.Valid(raw) {
t.Fatalf("expected valid strategy schema JSON")
}
}
func TestPlannerToolsKeepFullManageStrategyForMutationIntent(t *testing.T) {
tools := plannerToolsForText("创建一个 BTC 网格策略")
tool := findToolForTest(tools, "manage_strategy")
if tool == nil {
t.Fatalf("expected manage_strategy in strategy tools")
}
raw, _ := json.Marshal(tool.Function.Parameters)
if len(raw) < 1500 {
t.Fatalf("expected full strategy schema for mutation intent, got %d bytes", len(raw))
}
}
func toolNamesForTest(tools []mcp.Tool) []string {
names := make([]string, 0, len(tools))
for _, tool := range tools {
names = append(names, tool.Function.Name)
}
return names
}
func findToolForTest(tools []mcp.Tool, name string) *mcp.Tool {
for i := range tools {
if tools[i].Function.Name == name {
return &tools[i]
}
}
return nil
}