boocode/apps/server/src/services/inference/payload.ts

import type { Sql } from '../../db.js';
import type {
  Agent,
  Message,
  Project,
  Session,
} from '../../types/api.js';
import * as compaction from '../compaction.js';
import { buildSystemPrompt } from '../system-prompt.js';
import { isAnySentinel } from './sentinels.js';
import { PRUNE_TRIGGER_TOKENS, prune } from './prune.js';
import type { InferenceContext } from './turn.js';

export interface OpenAiMessage {
  role: 'system' | 'user' | 'assistant' | 'tool';
  content: string | null;
  tool_calls?: Array<{
    id: string;
    type: 'function';
    function: { name: string; arguments: string };
  }>;
  tool_call_id?: string;
  // v1.13.1-C: reasoning text from a prior assistant turn, sourced from
  // message_parts kind='reasoning' rows joined in via reasoning_parts on
  // the messages_with_parts view. stream-phase.ts/toModelMessages threads
  // this into the AI SDK ReasoningPart when forwarding to the model so
  // reasoning models can resume mid-thought across tool-call boundaries.
  reasoning?: string;
}

// v1.12: buildSystemPrompt lives in services/system-prompt.ts. It awaits the
// container-guidance loader, so this function is async too and every call
// site in inference.ts awaits the result.
export async function buildMessagesPayload(
  session: Session,
  project: Project,
  history: Message[],
  agent: Agent | null = null
): Promise<OpenAiMessage[]> {
  const out: OpenAiMessage[] = [];
  const systemPrompt = await buildSystemPrompt(project, session, agent);
  out.push({ role: 'system', content: systemPrompt });

  // Find the latest compact marker — only send messages from that point onwards
  let startIdx = 0;
  for (let i = history.length - 1; i >= 0; i--) {
    if (history[i]!.kind === 'compact') {
      startIdx = i;
      break;
    }
  }

  for (let i = startIdx; i < history.length; i++) {
    const m = history[i]!;
    if (m.kind === 'compact') {
      out.push({ role: 'system', content: m.content });
      continue;
    }
    // v1.8.2 / v1.11.6: cap-hit and doom-loop sentinels are UI-only — never
    // send them to the LLM. The synthetic instruction note lives only inside
    // the summary call's messages array and is never persisted, so on a
    // follow-up turn the model resumes with a clean context.
    if (isAnySentinel(m)) continue;
    if (m.role === 'assistant' && m.status === 'streaming') continue;
    if (m.role === 'assistant' && m.status === 'cancelled') continue;
    if (m.role === 'tool') {
      const tr = m.tool_results;
      if (!tr) continue;
      const outputText = tr.error
        ? `error: ${tr.error}`
        : typeof tr.output === 'string'
          ? tr.output
          : JSON.stringify(tr.output);
      out.push({
        role: 'tool',
        content: outputText,
        tool_call_id: tr.tool_call_id,
      });
      continue;
    }
    if (m.role === 'assistant') {
      const msg: OpenAiMessage = {
        role: 'assistant',
        content: m.content && m.content.length > 0 ? m.content : null,
      };
      if (m.tool_calls && m.tool_calls.length > 0) {
        msg.tool_calls = m.tool_calls.map((tc) => ({
          id: tc.id,
          type: 'function' as const,
          function: { name: tc.name, arguments: JSON.stringify(tc.args) },
        }));
      }
      // v1.13.1-C: collapse reasoning_parts into a single string. The view
      // returns them ordered by sequence; multiple reasoning parts on one
      // message are rare but concat preserves ordering. Skip when absent.
      if (m.reasoning_parts && m.reasoning_parts.length > 0) {
        msg.reasoning = m.reasoning_parts.map((p) => p.text ?? '').join('');
      }
      out.push(msg);
      continue;
    }
    out.push({ role: 'user', content: m.content });
  }
  return out;
}

export async function loadContext(
  sql: Sql,
  sessionId: string,
  chatId: string
): Promise<{ session: Session; project: Project; history: Message[] } | null> {
  const sessionRows = await sql<Session[]>`
    SELECT id, project_id, name, model, system_prompt, status, created_at, updated_at,
           agent_id, web_search_enabled
    FROM sessions WHERE id = ${sessionId}
  `;
  if (sessionRows.length === 0) return null;
  const session = sessionRows[0]!;

  const projectRows = await sql<Project[]>`
    SELECT id, name, path, added_at, last_session_id, status, gitea_remote,
           default_system_prompt, default_web_search_enabled
    FROM projects WHERE id = ${session.project_id}
  `;
  if (projectRows.length === 0) return null;
  const project = projectRows[0]!;

  // v1.11: filter compacted messages out of the inference assembly. The GET
  // /api/sessions/:id/messages endpoint still returns everything (so the UI
  // can show history with the summary card inline); only LLM payloads skip
  // compacted rows. compacted_at IS NULL keeps the active summary + tail.
  // v1.13.1-B: reads tool_calls/tool_results via the parts-merged view.
  // v1.13.1-C: also pull reasoning_parts so assistant messages from
  // reasoning models can be replayed with their reasoning context preserved.
  const history = await sql<Message[]>`
    SELECT id, session_id, chat_id, role, content, kind, tool_calls, tool_results, status, last_seq,
           tokens_used, ctx_used, ctx_max, started_at, finished_at, created_at, metadata,
           reasoning_parts
    FROM messages_with_parts
    WHERE chat_id = ${chatId} AND compacted_at IS NULL
    ORDER BY created_at ASC, id ASC
  `;

  return { session, project, history };
}

// v1.11: shared helper used after both finalizeCompletion and executeToolPhase
// persist their token counts. Reads tokens off the just-UPDATEd row (which
// the caller returns from RETURNING), runs compaction.isOverflow, and flips
// chats.needs_compaction. The next runAssistantTurn invocation acts on it.
// Silent on missing tokens — llama-swap occasionally omits usage on truncated
// streams, and we'd rather miss one overflow than crash the inference path.
export async function maybeFlagForCompaction(
  ctx: InferenceContext,
  chatId: string,
  updated: { tokens_used: number | null; ctx_used: number | null; ctx_max: number | null } | undefined,
): Promise<void> {
  if (!updated) return;
  const promptTokens = updated.ctx_used;
  const completionTokens = updated.tokens_used;
  const contextLimit = updated.ctx_max;
  if (typeof promptTokens !== 'number') return;
  if (typeof completionTokens !== 'number') return;
  if (typeof contextLimit !== 'number') return;
  const overflow = compaction.isOverflow(
    { prompt_tokens: promptTokens, completion_tokens: completionTokens },
    contextLimit,
  );
  if (!overflow) return;

  // v1.13.4: try the cheap prune first. If it freed at least the buffer
  // worth of tokens (PRUNE_TRIGGER_TOKENS, identical to COMPACTION_BUFFER),
  // we're below the threshold again — skip flagging summarize for the next
  // turn. The next turn's overflow check will re-evaluate from scratch.
  // Prune failures (DB errors etc.) propagate so the surrounding inference
  // path sees them; the catch in finalizeCompletion / executeToolPhase
  // doesn't shield this — by design, we want to know if prune is broken.
  const pruned = await prune({ sql: ctx.sql, chatId });
  if (pruned.hidden > 0) {
    ctx.log.info(
      { chatId, hidden: pruned.hidden, freedTokens: pruned.freedTokens },
      'inference: prune freed context budget',
    );
  }
  if (pruned.freedTokens >= PRUNE_TRIGGER_TOKENS) {
    // Prune handled it; skip the (expensive) summarize path.
    return;
  }

  await ctx.sql`UPDATE chats SET needs_compaction = true WHERE id = ${chatId}`;
  ctx.log.info({ chatId, promptTokens, completionTokens, contextLimit }, 'inference: flagged for compaction');
}