047e2a4070
- build() is now suspend: pipeline runs all stages in fixed order, each gated by level - TOKEN_PRUNE (level 3): TokenPruner interface + LLMLingua-2 sidecar (sidecars/llmlingua) + HttpTokenPruner adapter (fails open if sidecar down); prunes freeform, preserves protected spans, skips tier-0 turns when TIER_SPLIT on - TOME_MERGE (level 8): ToMeMerger collapses near-duplicate freeform turns (Jaccard) - Stage 5 selection: RelevanceScorer + EmbeddingRelevanceScorer (cosine over Embedder); query-conditioned reorder so least-relevant freeform drops first under budget - [orchestration] compression_level + token_pruner_url config, wired in Main - suspend ripple fixed across builder callers/stubs
124 lines
5.5 KiB
Kotlin
124 lines
5.5 KiB
Kotlin
import com.correx.core.context.builder.DefaultContextPackBuilder
|
|
import com.correx.core.context.compression.DefaultContextCompressor
|
|
import com.correx.core.context.model.ContextEntry
|
|
import com.correx.core.context.model.ContextLayer
|
|
import com.correx.core.context.model.ContextPack
|
|
import com.correx.core.context.model.EntryRole
|
|
import com.correx.core.context.model.TokenBudget
|
|
import com.correx.core.events.types.ContextEntryId
|
|
import com.correx.core.events.types.ContextPackId
|
|
import com.correx.core.events.types.SessionId
|
|
import com.correx.core.events.types.StageId
|
|
import com.correx.core.inference.PromptRenderer
|
|
import org.junit.jupiter.api.Assertions.assertEquals
|
|
import org.junit.jupiter.api.Test
|
|
|
|
class PromptRendererOrderingTest {
|
|
|
|
private val sessionId = SessionId("s1")
|
|
private val stageId = StageId("stage-1")
|
|
|
|
private fun entry(id: String, layer: ContextLayer, role: EntryRole, sourceType: String) = ContextEntry(
|
|
id = ContextEntryId(id),
|
|
layer = layer,
|
|
content = id,
|
|
sourceType = sourceType,
|
|
sourceId = id,
|
|
tokenEstimate = 10,
|
|
role = role,
|
|
)
|
|
|
|
@Test
|
|
fun `stage tool loop renders chronologically with the task before tool turns`() = kotlinx.coroutines.runBlocking {
|
|
// The builder stamps ordinals from input order; the renderer must honour them so the
|
|
// model sees: system, task, assistant call, tool result, assistant call, tool result.
|
|
val builder = DefaultContextPackBuilder(DefaultContextCompressor())
|
|
val entries = listOf(
|
|
entry("sys", ContextLayer.L0, EntryRole.SYSTEM, "systemPrompt"),
|
|
entry("task", ContextLayer.L1, EntryRole.USER, "agentPrompt"),
|
|
entry("call1", ContextLayer.L2, EntryRole.ASSISTANT, "assistantToolCall"),
|
|
entry("result1", ContextLayer.L2, EntryRole.TOOL, "toolResult"),
|
|
entry("call2", ContextLayer.L2, EntryRole.ASSISTANT, "assistantToolCall"),
|
|
entry("result2", ContextLayer.L2, EntryRole.TOOL, "toolResult"),
|
|
)
|
|
val pack = builder.build(ContextPackId("p"), sessionId, stageId, entries, TokenBudget(limit = 4000))
|
|
|
|
val messages = PromptRenderer.render(pack)
|
|
assertEquals(
|
|
listOf("system" to "sys", "user" to "task", "assistant" to "call1",
|
|
"tool" to "result1", "assistant" to "call2", "tool" to "result2"),
|
|
messages.map { it.role to it.content },
|
|
)
|
|
Unit
|
|
}
|
|
|
|
@Test
|
|
fun `without ordinals the live user turn still renders last (router chat fallback)`() {
|
|
// Packs built outside DefaultContextPackBuilder (router chat) leave ordinal at 0;
|
|
// the renderer falls back to layer priority so L1 (the user query) lands last.
|
|
val pack = ContextPack(
|
|
id = ContextPackId("p"),
|
|
sessionId = sessionId,
|
|
stageId = stageId,
|
|
layers = mapOf(
|
|
ContextLayer.L1 to listOf(entry("question", ContextLayer.L1, EntryRole.USER, "chat")),
|
|
ContextLayer.L2 to listOf(entry("memory", ContextLayer.L2, EntryRole.SYSTEM, "l2")),
|
|
),
|
|
budgetUsed = 20,
|
|
budgetLimit = 4000,
|
|
)
|
|
val messages = PromptRenderer.render(pack)
|
|
assertEquals(listOf("system" to "memory", "user" to "question"), messages.map { it.role to it.content })
|
|
}
|
|
|
|
@Test
|
|
fun `steering directives are re-anchored as a trailing user reminder`() {
|
|
// A steering note folds into the leading system block AND is restated at the tail so a
|
|
// weak local model still sees the active constraint next to the final query.
|
|
val pack = ContextPack(
|
|
id = ContextPackId("p"),
|
|
sessionId = sessionId,
|
|
stageId = stageId,
|
|
layers = mapOf(
|
|
ContextLayer.L0 to listOf(entry("sys", ContextLayer.L0, EntryRole.SYSTEM, "systemPrompt")),
|
|
ContextLayer.L2 to listOf(entry("directive", ContextLayer.L2, EntryRole.SYSTEM, "steeringNote")),
|
|
ContextLayer.L1 to listOf(entry("question", ContextLayer.L1, EntryRole.USER, "chat")),
|
|
),
|
|
budgetUsed = 30,
|
|
budgetLimit = 4000,
|
|
)
|
|
val messages = PromptRenderer.render(pack)
|
|
assertEquals(
|
|
listOf(
|
|
"system" to "sys\n\ndirective",
|
|
"user" to "question",
|
|
"user" to "Reminder — active steering directive(s):\ndirective",
|
|
),
|
|
messages.map { it.role to it.content },
|
|
)
|
|
}
|
|
|
|
@Test
|
|
fun `non-L0 SYSTEM entries fold into the single leading system message`() {
|
|
// Strict chat templates (Qwen) 400 on any system message after index 0. Recalled L3
|
|
// memory and L2 summaries carry EntryRole.SYSTEM and must merge into the system block.
|
|
val pack = ContextPack(
|
|
id = ContextPackId("p"),
|
|
sessionId = sessionId,
|
|
stageId = stageId,
|
|
layers = mapOf(
|
|
ContextLayer.L0 to listOf(entry("sys", ContextLayer.L0, EntryRole.SYSTEM, "systemPrompt")),
|
|
ContextLayer.L1 to listOf(entry("question", ContextLayer.L1, EntryRole.USER, "chat")),
|
|
ContextLayer.L3 to listOf(entry("recalled", ContextLayer.L3, EntryRole.SYSTEM, "recalledMemory")),
|
|
),
|
|
budgetUsed = 30,
|
|
budgetLimit = 4000,
|
|
)
|
|
val messages = PromptRenderer.render(pack)
|
|
assertEquals(
|
|
listOf("system" to "sys\n\nrecalled", "user" to "question"),
|
|
messages.map { it.role to it.content },
|
|
)
|
|
}
|
|
}
|