soraacode 0.2.0 → 0.2.1

package/src/session/llm.ts

@@ -1,390 +0,0 @@
-import { LayerNode } from "@soraacode/core/effect/layer-node"
-import { llmClient } from "@soraacode/core/effect/app-node-platform"
-import { PermissionV1 } from "@soraacode/core/v1/permission"
-import { Provider } from "@/provider/provider"
-import { SessionV1 } from "@soraacode/core/v1/session"
-import { serviceUse } from "@soraacode/core/effect/service-use"
-import { Context, Effect, Layer } from "effect"
-import * as Stream from "effect/Stream"
-import { streamText, wrapLanguageModel, type ModelMessage, type Tool } from "ai"
-import type { LLMEvent } from "@soraacode/llm"
-import { LLMClient } from "@soraacode/llm/route"
-import type { LLMClientService } from "@soraacode/llm/route"
-import { GitLabWorkflowLanguageModel } from "gitlab-ai-provider"
-import { ProviderTransform } from "@/provider/transform"
-import { Config } from "@/config/config"
-import type { Agent } from "@/agent/agent"
-import type { MessageV2 } from "./message-v2"
-import { Plugin } from "@/plugin"
-import { Permission } from "@/permission"
-import { EventV2Bridge } from "@/event-v2-bridge"
-import { EventV2 } from "@soraacode/core/event"
-import { Wildcard } from "@/util/wildcard"
-import { SessionID } from "@/session/schema"
-import { Auth } from "@/auth"
-import { EffectBridge } from "@/effect/bridge"
-import { RuntimeFlags } from "@/effect/runtime-flags"
-import * as Option from "effect/Option"
-import { LLMAISDK } from "./llm/ai-sdk"
-import { LLMNativeRuntime } from "./llm/native-runtime"
-import { LLMRequestPrep } from "./llm/request"
-
-export const OUTPUT_TOKEN_MAX = ProviderTransform.OUTPUT_TOKEN_MAX
-
-export type StreamInput = {
-  user: SessionV1.User
-  sessionID: string
-  parentSessionID?: string
-  model: Provider.Model
-  agent: Agent.Info
-  permission?: PermissionV1.Ruleset
-  system: string[]
-  messages: ModelMessage[]
-  small?: boolean
-  tools: Record<string, Tool>
-  retries?: number
-  toolChoice?: "auto" | "required" | "none"
-}
-
-export type StreamRequest = StreamInput & {
-  abort: AbortSignal
-}
-
-export interface Interface {
-  readonly stream: (input: StreamInput) => Stream.Stream<LLMEvent, unknown>
-}
-
-export class Service extends Context.Service<Service, Interface>()("@opencode/LLM") {}
-
-export const use = serviceUse(Service)
-
-const live: Layer.Layer<
-  Service,
-  never,
-  | Auth.Service
-  | Config.Service
-  | Provider.Service
-  | Plugin.Service
-  | Permission.Service
-  | EventV2Bridge.Service
-  | LLMClientService
-  | RuntimeFlags.Service
-> = Layer.effect(
-  Service,
-  Effect.gen(function* () {
-    const auth = yield* Auth.Service
-    const config = yield* Config.Service
-    const provider = yield* Provider.Service
-    const plugin = yield* Plugin.Service
-    const perm = yield* Permission.Service
-    const events = yield* EventV2Bridge.Service
-    const llmClient = yield* LLMClient.Service
-    const flags = yield* RuntimeFlags.Service
-
-    const run = Effect.fn("LLM.run")(function* (input: StreamRequest) {
-      yield* Effect.logInfo("stream", {
-        providerID: input.model.providerID,
-        modelID: input.model.id,
-        "session.id": input.sessionID,
-        small: (input.small ?? false).toString(),
-        agent: input.agent.name,
-        mode: input.agent.mode,
-      })
-
-      const [language, cfg, item, info] = yield* Effect.all(
-        [
-          provider.getLanguage(input.model),
-          config.get(),
-          provider.getProvider(input.model.providerID),
-          auth.get(input.model.providerID),
-        ],
-        { concurrency: "unbounded" },
-      )
-
-      const isWorkflow = language instanceof GitLabWorkflowLanguageModel
-      const prepared = yield* LLMRequestPrep.prepare({
-        ...input,
-        provider: item,
-        auth: info,
-        plugin,
-        flags,
-        isWorkflow,
-      })
-
-      // Wire up toolExecutor for DWS workflow models so that tool calls
-      // from the workflow service are executed via opencode's tool system
-      // and results sent back over the WebSocket.
-      const bridge = yield* EffectBridge.make()
-      if (language instanceof GitLabWorkflowLanguageModel) {
-        const workflowModel = language as GitLabWorkflowLanguageModel & {
-          sessionID?: string
-          sessionPreapprovedTools?: string[]
-          approvalHandler?: (approvalTools: { name: string; args: string }[]) => Promise<{ approved: boolean }>
-        }
-        workflowModel.sessionID = input.sessionID
-        workflowModel.systemPrompt = prepared.system.join("\n")
-        workflowModel.toolExecutor = async (toolName, argsJson, _requestID) => {
-          const t = prepared.tools[toolName]
-          if (!t || !t.execute) {
-            return { result: "", error: `Unknown tool: ${toolName}` }
-          }
-          try {
-            const result = await t.execute!(JSON.parse(argsJson), {
-              toolCallId: _requestID,
-              messages: input.messages,
-              abortSignal: input.abort,
-            })
-            const output = typeof result === "string" ? result : (result?.output ?? JSON.stringify(result))
-            return {
-              result: output,
-              metadata: typeof result === "object" ? result?.metadata : undefined,
-              title: typeof result === "object" ? result?.title : undefined,
-            }
-          } catch (e: any) {
-            return { result: "", error: e.message ?? String(e) }
-          }
-        }
-
-        const ruleset = Permission.merge(input.agent.permission ?? [], input.permission ?? [])
-        workflowModel.sessionPreapprovedTools = Object.keys(prepared.tools).filter((name) => {
-          const match = ruleset.findLast((rule) => Wildcard.match(name, rule.permission))
-          return !match || match.action !== "ask"
-        })
-
-        const approvedToolsForSession = new Set<string>()
-        workflowModel.approvalHandler = bridge.bind(async (approvalTools) => {
-          const uniqueNames = [...new Set(approvalTools.map((t: { name: string }) => t.name))] as string[]
-          // Auto-approve tools that were already approved in this session
-          // (prevents infinite approval loops for server-side MCP tools)
-          if (uniqueNames.every((name) => approvedToolsForSession.has(name))) {
-            return { approved: true }
-          }
-
-          const id = PermissionV1.ID.ascending()
-          let unsub: EventV2.Unsubscribe | undefined
-          try {
-            unsub = await bridge.promise(
-              events.listen((event) => {
-                if (event.type !== Permission.Event.Replied.type) return Effect.void
-                const data = event.data as EventV2.Data<typeof Permission.Event.Replied>
-                if (data.requestID !== id) return Effect.void
-                void data.reply
-                return Effect.void
-              }),
-            )
-            const toolPatterns = approvalTools.map((t: { name: string; args: string }) => {
-              try {
-                const parsed = JSON.parse(t.args) as Record<string, unknown>
-                const title = (parsed?.title ?? parsed?.name ?? "") as string
-                return title ? `${t.name}: ${title}` : t.name
-              } catch {
-                return t.name
-              }
-            })
-            const uniquePatterns = [...new Set(toolPatterns)] as string[]
-            await bridge.promise(
-              perm.ask({
-                id,
-                sessionID: SessionID.make(input.sessionID),
-                permission: "workflow_tool_approval",
-                patterns: uniquePatterns,
-                metadata: { tools: approvalTools },
-                always: uniquePatterns,
-                ruleset: [],
-              }),
-            )
-            for (const name of uniqueNames) approvedToolsForSession.add(name)
-            workflowModel.sessionPreapprovedTools = [...(workflowModel.sessionPreapprovedTools ?? []), ...uniqueNames]
-            return { approved: true }
-          } catch {
-            return { approved: false }
-          } finally {
-            if (unsub) await bridge.promise(unsub)
-          }
-        })
-      }
-
-      const tracer = undefined
-      const telemetryTracer = undefined
-
-      // Runtime seam: native is an opt-in adapter over @soraacode/llm. It
-      // either returns a ready LLMEvent stream or a concrete fallback reason.
-      if (flags.experimentalNativeLlm) {
-        const native = LLMNativeRuntime.stream({
-          model: input.model,
-          provider: item,
-          auth: info,
-          llmClient,
-          messages: prepared.messages,
-          tools: prepared.tools,
-          toolChoice: input.toolChoice,
-          temperature: prepared.params.temperature,
-          topP: prepared.params.topP,
-          topK: prepared.params.topK,
-          maxOutputTokens: prepared.params.maxOutputTokens,
-          providerOptions: prepared.params.options,
-          headers: prepared.headers,
-          abort: input.abort,
-        })
-        if (native.type === "supported") {
-          yield* Effect.logInfo("llm runtime selected", {
-            "llm.runtime": "native",
-            "llm.provider": input.model.providerID,
-            "llm.model": input.model.id,
-          })
-          return {
-            type: "native" as const,
-            stream: native.stream,
-          }
-        }
-        yield* Effect.logInfo("llm runtime selected", {
-          "llm.runtime": "ai-sdk",
-          "llm.provider": input.model.providerID,
-          "llm.model": input.model.id,
-          "llm.native_unsupported_reason": native.reason,
-        })
-        yield* Effect.logInfo("native runtime unavailable; falling back to ai-sdk", {
-          providerID: input.model.providerID,
-          modelID: input.model.id,
-          "session.id": input.sessionID,
-          small: (input.small ?? false).toString(),
-          agent: input.agent.name,
-          mode: input.agent.mode,
-          reason: native.reason,
-        })
-      }
-
-      yield* Effect.logInfo("llm runtime selected", {
-        "llm.runtime": "ai-sdk",
-        "llm.provider": input.model.providerID,
-        "llm.model": input.model.id,
-      })
-      // Default runtime path: AI SDK owns provider execution and tool dispatch;
-      // LLMAISDK.toLLMEvents below normalizes fullStream parts for the processor.
-      return {
-        type: "ai-sdk" as const,
-        result: streamText({
-          onError(error) {
-            bridge.fork(
-              Effect.logError("stream error", {
-                providerID: input.model.providerID,
-                modelID: input.model.id,
-                "session.id": input.sessionID,
-                small: (input.small ?? false).toString(),
-                agent: input.agent.name,
-                mode: input.agent.mode,
-                error,
-              }),
-            )
-          },
-          // Copilot returns the authoritative billed amount only in provider-specific response fields.
-          includeRawChunks: input.model.providerID.includes("github-copilot"),
-          async experimental_repairToolCall(failed) {
-            const lower = failed.toolCall.toolName.toLowerCase()
-            if (lower !== failed.toolCall.toolName && prepared.tools[lower]) {
-              return {
-                ...failed.toolCall,
-                toolName: lower,
-              }
-            }
-            return {
-              ...failed.toolCall,
-              input: JSON.stringify({
-                tool: failed.toolCall.toolName,
-                error: failed.error.message,
-              }),
-              toolName: "invalid",
-            }
-          },
-          temperature: prepared.params.temperature,
-          topP: prepared.params.topP,
-          topK: prepared.params.topK,
-          providerOptions: ProviderTransform.providerOptions(input.model, prepared.params.options),
-          activeTools: Object.keys(prepared.tools).filter((x) => x !== "invalid"),
-          tools: prepared.tools,
-          toolChoice: input.toolChoice,
-          maxOutputTokens: prepared.params.maxOutputTokens,
-          abortSignal: input.abort,
-          headers: prepared.headers,
-          maxRetries: input.retries ?? 0,
-          messages: prepared.messages,
-          model: wrapLanguageModel({
-            model: language,
-            middleware: [
-              {
-                specificationVersion: "v3" as const,
-                async transformParams(args) {
-                  if (args.type === "stream") {
-                    // @ts-expect-error
-                    args.params.prompt = ProviderTransform.message(
-                      args.params.prompt,
-                      input.model,
-                      prepared.messageTransformOptions,
-                    )
-                  }
-                  return args.params
-                },
-              },
-            ],
-          }),
-          experimental_telemetry: {
-            isEnabled: cfg.experimental?.openTelemetry,
-            functionId: "session.llm",
-            tracer: telemetryTracer,
-            metadata: {
-              userId: cfg.username ?? "unknown",
-              sessionId: input.sessionID,
-            },
-          },
-        }),
-      }
-    })
-
-    const stream: Interface["stream"] = (input) =>
-      Stream.scoped(
-        Stream.unwrap(
-          Effect.gen(function* () {
-            const ctrl = yield* Effect.acquireRelease(
-              Effect.sync(() => new AbortController()),
-              (ctrl) => Effect.sync(() => ctrl.abort()),
-            )
-
-            const result = yield* run({ ...input, abort: ctrl.signal })
-
-            if (result.type === "native") return result.stream
-
-            // Adapter seam: both runtimes expose the same LLMEvent stream. Native
-            // already returns one; AI SDK streams are converted here.
-            const state = LLMAISDK.adapterState()
-            return Stream.fromAsyncIterable(result.result.fullStream, (e) =>
-              e instanceof Error ? e : new Error(String(e)),
-            ).pipe(
-              Stream.mapEffect((event) => LLMAISDK.toLLMEvents(state, event)),
-              Stream.flatMap((events) => Stream.fromIterable(events)),
-            )
-          }),
-        ),
-      )
-
-    return Service.of({ stream })
-  }),
-)
-
-export const hasToolCalls = LLMRequestPrep.hasToolCalls
-
-export const node = LayerNode.make({
-  service: Service,
-  layer: live,
-  deps: [
-    Auth.node,
-    Config.node,
-    Provider.node,
-    Plugin.node,
-    Permission.node,
-    EventV2Bridge.node,
-    llmClient,
-    RuntimeFlags.node,
-  ],
-})
-
-export * as LLM from "./llm"

package/src/session/message-error.ts

@@ -1,14 +0,0 @@
-import { Schema } from "effect"
-import { NamedError } from "@soraacode/core/util/error"
-
-export const OutputLengthError = NamedError.create("MessageOutputLengthError", {})
-
-export const AuthError = NamedError.create("ProviderAuthError", {
-  providerID: Schema.String,
-  message: Schema.String,
-})
-
-export const Shared = [AuthError.EffectSchema, NamedError.Unknown.EffectSchema, OutputLengthError.EffectSchema] as const
-export const SharedSchema = Schema.Union(Shared)
-
-export * as MessageError from "./message-error"