Signaller logoSignaller

Instruction

Full instruction for Signaller: sources, prompt, images, auto-posting and billing.

Other instructions
🧠Prompt Engineering📚Prompt Library

Signaller Instruction

Quick start

  1. Add sources using /add.
  2. Wait for the bot to deliver posts.

Who is this for?

  • Channel owners — run your channel fully or partially with AI (text generation, images, auto‑posting).
  • People searching for clients/leads in real time — monitor keywords in chats and channels.
  • Anyone who needs live updates from Telegram and X (Twitter) with automatic filtering/processing.

Step‑by‑step scenarios

Running a Telegram channel with AI

Maximize automation for content creation.

  1. Add sources: find quality Telegram or X accounts in your niche and add via /add. Search in Telegram, Google/Yandex, and aggregators like tgstat.ru.
  2. Define your voice: write a detailed /prompt for the AI.
  3. Configure images:
    • Full automation (auto‑posting): enable auto image generation in /settings. Safer for copyrights.
    • Manual curation: enable Google image search in /settings (review copyrights).
  4. Connect your channel: add the bot as an admin and send the link via /channel.
  5. Enable auto‑posting (optional): turn it on in /settings.
  6. Optimize cadence (optional): set a minimal interval (e.g., 240 minutes) and duplicate checks in /settings.
  7. (Optional) Watermark: configure via /watermark. Applied on publish.
  8. (Optional) Multiple channels/teamwork: create chats, add the bot as admin — settings are unique per chat.

Finding clients and leads

This scenario turns the bot into your lead‑generation assistant. No publishing channel required.

  1. Add sources: via /add include chats and channels with potential clients (Telegram search, Google/Yandex, tgstat.ru).
  2. Set strict filtering: The key tool is /prompt. Define keywords and selection rules; when not relevant, the AI should output only “no signal”.
  3. Smart forward (optional): enable in /settings to receive original messages. Note users can delete their messages — sometimes it’s better to keep this off.
  4. Receive notifications: the bot sends filtered matches to your DM.
  5. (Optional) Teamwork: create a chat, add the bot as admin, and share the signal flow.

Filtering and processing information

For personal use: keep what matters and optionally summarize.

  1. Add sources: via /add.
  2. Define processing rules: In /prompt specify the logic. Examples:
    • Example 1 (filter only): Write "no signal" if the post is about politics or ads. Otherwise send only the original post text.
    • Example 2 (filter + processing): If the post is about new technologies, write a 2–3 sentence summary. Otherwise output "no signal".
  3. Choose delivery:
    • Only original filtered posts — enable Smart forward in /settings.
    • AI‑processed posts — keep Smart forward off.
  4. (Optional) Publish to a channel: configure via /channel.
  5. (Optional) Teamwork: use chats with the bot as admin.

Feature reference

Adding sources to monitor

  • Command: /add
  • Supported formats: https://x.com/Tesla, https://t.me/telegram, telegram / @telegram, 1005640892.
  • For private channels requiring approval, contact support @signaller_support.
  • Where to find sources: Telegram search, tgstat.ru, etc.
  • Update speed: Telegram — instant; X — every ~15 minutes.

Prompt configuration

The prompt is your instruction for the AI. Detailed article: prompt engineering.

  • Command: /prompt
  • The AI sees only the first image in a post.
  • Key filtering phrase: no signal (or Russian нет сигнала). Specify when to output it.
  • Formatting: HTML tags supported: <b>, <a>, <strong>, <i>, <em>, <code>, <s>, <strike>, <del>, <u>, <pre>, <blockquote>, <tg-spoiler>.

Service workflow

No signal — Skip
News

From source

➝
AI

Processing

➝
optional Review

By user

➝
Publish

To your channel

Available HTML tags

  • <b>bold</b>, <strong>bold</strong>
  • <i>italic</i>, <em>italic</em>
  • <code>code</code>
  • <s>strikethrough</s>, <strike>strikethrough</strike>, <del>strikethrough</del>
  • <u>underlined</u>
  • <pre language="c++">code block</pre> 
    example of a code block
  • <a href="https://t.me/signaller_ai_bot">link</a>
    example link
  • <blockquote>quote</blockquote>
    quote example
  • <tg-spoiler>spoiler</tg-spoiler>

Suggested prompt template:

Your role: <e.g., technology analyst>
Your task: <e.g., rephrase the text>
Post formatting: <use HTML, emojis, style>
Filtering: output “no signal” if <conditions>

The AI receives a single concatenated text like this:

<your prompt>

POST METADATA:
<author; channel;>

POST TEXT:
<incoming post>

Language models

  • gemini-2.0-flash-lite — ultra‑fast and cheap; mass pre‑processing and quick filtering.
  • gemini-2.0-flash — balance of speed and quality for small context.
  • gemini-2.5-flash-lite — economical 2.5 version for batch processing.
  • gemini-2.5-flash — stronger at context while still fast.
  • gemini-2.5-pro — flagship for complex analysis and multimodality.
  • gpt-5 — heavy tool for code and deep reasoning.
  • gpt-5-mini — economical and versatile.
  • gpt-5-nano — fast and cheap for routine tasks.

Image generation models

Test prompt (to compare models):

A cat‑knight in shining engraved armor holding a tiny sword and shield, standing atop a stone castle tower, sunset and fiery sky in the background, epic fantasy atmosphere.

Results:

Flux Schnell — budget, fastblack-forest-labs/flux-schnell — Flux Schnell — budget, fast
Flux 1.1 Pro — high qualityblack-forest-labs/flux-1.1-pro — Flux 1.1 Pro — high quality
Flux 1.1 Pro Ultra — ultra qualityblack-forest-labs/flux-1.1-pro-ultra — Flux 1.1 Pro Ultra — ultra quality
Imagen 4 Fast — economy modegoogle/imagen-4-fast — Imagen 4 Fast — economy mode
Imagen 4 — main modelgoogle/imagen-4 — Imagen 4 — main model
Imagen 4 Ultra — premiumgoogle/imagen-4-ultra — Imagen 4 Ultra — premium quality

Smart forward mode

  • Enable in /settings. Lets you receive original messages that passed your prompt filter without generating new text.
  • Example prompt for forwarding with filtering only:
Send me only posts about carrots, no other vegetables, fruits, or items.

If a post is not about carrots, output only "no signal".

If it is about carrots, output only "yes".
  • This mode saves tokens by skipping language model outputs.

Minimal interval filter

  • Command: /settings
  • Skips posts when less than the specified time has passed since the last publish (e.g., 240 minutes).
  • You don’t pay for posts skipped by this interval.

Duplicate check

  • Command: /settings
  • (Experimental) Skips semantically similar posts within the last 24 hours.
  • You don’t pay for posts skipped as duplicates.

Working with images

Auto image generation and Google image search are available only for Telegram messages. For X, tweet screenshots are created automatically (you can generate an image manually if auto‑posting is off).

1) Auto image generation

  • Enable in /settings.
  • /image_models — choose a model; /image_prompt — configure a prompt (in English) that your LLM uses to generate the image model’s prompt.
  • Only generated images are allowed for auto‑posting.
Example image generated by flux-schnellExample image generated by black-forest-labs/flux-schnell
  • Enable in /settings.
  • New found images include a source link — review copyrights and possible watermarks.
  • Auto‑posting is disabled for Google images; manual review is required.
Example post with Google imageExample post with a Google image

3) Watermark

  • Command: /watermark
  • Adds a small text in the image corner when publishing to a channel.
  • Not shown in review previews but added on send.
Watermark exampleWatermark example

Publishing posts

  1. Channel setup:

    • Add the bot as admin.
    • Send /channel and your channel link.
    • Confirm by a channel admin.

    1.1. Add the bot as an administrator to the publishing channel.

    Adding the bot to the channelAdding the bot to the channel

    1.2. Copy the channel link (e.g., https://t.me/telegram or @telegram).

    1.3. Use /channel and send the link.

    1.4. Confirm the addition — only a channel admin can do this.

    After this you’ll see buttons to send posts to the channel.

  2. Auto‑posting:

    • After setup, enable auto‑posting in /settings. Posts will be published automatically.

  3. Manual posting:

    • “Send post” buttons are available after channel setup.
    • “Edit” lets you change the text and attach your image/video (up to 200 MB).

  4. Forwarding without “forwarded from…“:

    • When forwarding manually, click “Forward settings” in Telegram and select “Hide sender name”. Inline buttons are not sent when forwarding.
    Forwarded from signaller“forwarded from signaller” label
    2147483648-214391Forward settings button
    Hide sender nameHide sender name

Post history

  • Command: /posts_history
  • Shows posts filtered by “no signal”, duplicates, and interval, plus processing details.

Teamwork and multiple channels

  • Add the bot as admin to a group chat.
  • Settings are unique per chat (/prompt, /add, /watermark, /settings, …).
  • Billing is tied to the user who first added the bot to the chat.

Billing & subscription

  • Command: /profile — balance, statements, token purchases.
  • Costs (summary):
X monitoring: 9 tokens per check (every ~15 minutes).

Google image search: 26 tokens per search.

Gemini processing: ~1–15 tokens (depends on model and text/image length). You also pay for posts filtered with “no signal”.

Image generation: from ~180 tokens (black-forest-labs/flux-schnell) to ~3600 (black-forest-labs/flux-1.1-pro-ultra) + LLM.
  • Free: interval and duplicate filters.

Feedback

Contact us: @signaller_support