Dashboard

Overview

0

Total Configured

0

Active Tools

0

Public Visibility

0

Auto-Refreshable

0 items
Function Description Method Active Public Version / Sync Action

Sync Status

What is currently live on the bot, and which tools have new unsynced changes.

Tool Method Bot version Latest Status

Pending Reviews

Select a change on the left to inspect its diff. Click a line number to comment.

No pending reviews

All tool changes are approved or review workflow is disabled.

Branches

Work on feature branches, then open a pull request to merge into main (Bitbucket-style).

Pull Requests

Review diffs, approve, and merge feature branches into main. You cannot approve your own PR.

Relationship Graph

AI-discovered dependencies and relationships between the tools of this environment. گراف ارتباطات بین ابزارها

AI Profiles
Tool No relationships Inactive Explicit (directly evident) Hidden (inferred) Click a node to highlight paths · scroll to zoom · drag to pan

Holistic Tool Optimization

Improve all tool descriptions together, respecting dependencies from the saved relationship graph. Approve changes to register each as a new version. بهینه‌سازی یکپارچه توضیحات با در نظر گرفتن وابستگی‌ها

AI Profiles

Select AI profile(s) and click Generate Suggestions to rewrite all active tool descriptions as a coherent set. Generate the relationship graph first for best dependency context.

Assistant thread QC

Fetch production PodChat threads, then run QC on a time range with a chosen AI profile.

SDK LLM

Fetch threads

Fetch pulls only messages after the last fetch. Refetch lists every PodChat thread to the last page, then pulls full message history; duplicates are skipped.

Last fetch: —

0 threads · 0 messages

Ready to fetch from PodChat.

Stored threads

    Select a thread

    Fetch threads, then pick one to read the conversation.

    Live PodChat stream

    Connects to the production socket and shows threads/messages as they arrive.

    Disconnected.

    Offline

    Live threads

      Select a thread

      Connect to the socket, then pick a thread to watch messages live.

      QC runs on the server. You can leave this page; progress continues and the finished report stays in history. Production tools and the Bot Integration system prompt are included automatically.

      No QC run in progress.

      Runs

        Thread results

          Select a QC result

          Pick a thread from the list to inspect findings.

          Each completed QC run stores a metrics dashboard in history. Compare two runs (for example the same time range before and after a prompt change) to see which indicators improved or got worse.

          Report history

            Run QC, then open a report from history to see the dashboard.

            Create Function

            Configure endpoint details and LLM context.

            Name & Description

            How the model identifies and understands this function.

            Use ### Section title headings matching Settings → Description Structure for consistent tools across the catalog.

            Endpoint & Schema

            HTTP endpoint, parameters, and response mapping.

            Parameters (JSON Schema)
            No arguments defined. The function accepts zero parameters.
            Response output (SelectedFieldsFromResult)

            Token & Constants

            Registry auth token type and fixed name/value pairs.

            Const Arguments

            Fixed name/value pairs sent to Tool Registry (not part of the LLM schema).

            No const arguments.

            Review & Save

            Confirm settings before saving.

            Execution flags

            Active (Included in Sync)
            Public Access
            Token Refreshable

            Authentication

            Global token and Bearer Prefix configurations.

            This unified token handles all API operations. (Can also accept OAuth redirect flow)

            'Bearer' Prefix Control

            Select which services should automatically have the `Bearer ` prefix attached to the token during API calls.

            Tool Management APIs (CRUD)
            Auth Refresh API
            Auth Code Exchange API
            Bot Sync API (Deployments)

            Core Setup

            Primary API endpoints configuration.

            Must end at the Core gateway root (paths like /api/management/tools are appended automatically).

            Workflow

            Choose how tool changes reach main and deploy.

            Like Bitbucket: edit on branches, review via PR, merge to main, then deploy.

            Bot Integration

            Configure the target LLM orchestrator and dynamic payloads.

            LLM Options

            Must be a valid JSON object. Sent exactly as written to the bot integration payload.

            Context Policy

            Controls how prior bot runs are included in the LLM context. Sent to Bot Provider as contextPolicyJson.

            Strategy · CUSTOM
            Recent Runs

            چه مقدار از Runهای سابق به LLM ارسال شود. فیلتر تعداد و بازه زمانی با هم اعمال می‌شوند — مثلاً ۱۰ Run آخر در ۱ ساعت؛ فقط Runهایی که داخل همان بازه باشند (تا سقف Count) می‌روند.

            Summary

            خلاصه‌سازی Runهای قبلی. اگر غیرفعال باشد ارسال نمی‌شود. Count یعنی تعداد Runهایی که برای ساخت خلاصه استفاده می‌شوند.

            Payload Aggregation

            Add arbitrary key/values to inject into the root of the PATCH request.

            No extra payload fields defined.

            AI Profiles

            Reusable OpenAI-compatible models used by description optimization, holistic optimization, and the relationship graph. Pick one or several at each AI step to compare results. پروفایل‌های هوش مصنوعی

            NameModelDefaultSelectable by othersActions

            Description Structure

            Define ordered paragraph titles used when a tool uses structured descriptions. Empty paragraphs are omitted from the bot payload.

            Paragraphs

            Each row becomes a ### heading in the compiled markdown description.

            No paragraphs defined yet.

            Chat SDK

            PodChat connection used to fetch assistant threads for QC. Socket is locked to production.

            Fixed to production PodChat.

            Server hosts

            Last successful fetch: —

            QC LLM

            QC uses the existing AI Profiles for URL, key, and model. Pick the profile when you run QC. This page is only the QC prompt, thinking, and payload limits.

            Parallel threads speed up QC without changing the prompt. Delay is only used when parallelism is 1.

            When on, QC requests send thinking: enabled and reasoning_effort. DeepSeek-style APIs receive an explicit disabled when this is off.

            Read live from production Bot Integration on every QC run. Edit it under Bot Integration.

            Production tools are attached automatically at QC time.

            QC Schedule

            Backend jobs sync chat every minute, then run QC at 08:00 Iran time for the past 24 hours, publish a PDF, and SMS the link.

            Chat sync

            Last sync: —

            Backfill re-fetches threads in the lookback window. If a user is still missing, use Fetch by mobile — PodChat can find that thread by cellphoneNumber even when it is not in the recent list.

            Daily QC

            Last daily QC: —

            SMS (sms.ir)

            Used in the SMS link. Must be reachable from phones.

            Last SMS: —

            User Management

            Create users and control their access level.

            UsernameRoleCreatedActions

            Audit Log

            Every change is recorded: who, when and what.

            WhenUserActionEntityChanges