LINE Chatbot Blueprint: Student Setup and Testing Guide (Fix LINE Integration Issue in Dialogflow)

How the scenario works

The blueprint gives you a finished Make.com scenario: you only connect your own accounts and paste your own token. LINE talks to Make.com, Make.com asks your Dialogflow agent which intent matched, and one of three modules sends the reply.

Matched intents go down Route A, which sends either the intent’s text or its LINE custom payloads. Messages with no matching intent go down Route B, where Gemini writes the answer.

Before you import

This blueprint replaces Dialogflow’s built-in LINE integration, which fails with IamPermissionDeniedException on Google’s side.

LINE sends messages to Make.com, Make.com asks Dialogflow for the intent, and Make.com sends the reply. Your intents still come from your own Dialogflow agent.

Collect these values first and keep them in a private note. Never paste them into the group chat or onto slides.

What	Where to find it	Used in module
Channel access token (long-lived)	LINE Developers → your channel → Messaging API tab → Issue	7, 23, and the LINE connection in 35
Your user ID	LINE Developers → your channel → Basic settings tab	LINE connection in 35
Google Cloud project ID of your agent	Dialogflow console → gear icon → General → Google Project	3
OAuth Client ID and Client secret	Created in Step 3	Google connection in 3
Gemini API key	Google AI Studio → Get API key	Gemini connection in 34
Make.com account (free plan works)	make.com	Everything

Your agent should already have at least 5 intents that answer correctly in Dialogflow’s Try it now panel. If the red IamPermissionDeniedException banner appears in the Dialogflow console, click OK and keep editing. Do not use Integrations → LINE.

Step 1: Import the blueprint

1. Download LINE_chatbot_student_template_v2.blueprint.json.
2. In Make.com, open Scenarios → Create a new scenario.

3. Click the ⋯ menu at the bottom of the editor → Import blueprint → choose the file → Import blueprint.

4. Rename the scenario at the top left, for example LINE bot.

After import, several modules show a red warning. That is expected: the file carries no webhook, no connections and no tokens, so you add your own in the next steps. Module numbers in this guide match the small grey number under each module.

Module	Name in the editor	What you must do
1	LINE webhook	Create a new webhook (Step 2)
3	Detect an Intent	Create a Google connection, pick your project (Step 3)
7	LINE reply: Dialogflow answer	Paste your channel access token (Step 4)
23	LINE reply: rich payload	Paste your channel access token (Step 4)
34	Gemini answer	Create a Gemini connection, edit the prompt (Step 5)
35	LINE reply: Gemini answer	Create a LINE connection (Step 5)
2, 4, 5, 6, 19 to 22	Iterator, variables, routers, JSON helpers	Nothing; leave as they are

Step 2: Webhook (module 1) and LINE Developers

Module 1 is where LINE delivers every message. Each group needs its own webhook address.

1. Click module 1 LINE webhook → next to Webhook click Add → name it LINE <group name> → Save.

2. Click Copy address to clipboard. It looks like https://hook.us2.make.com/xxxxxxxx.

3. Open LINE Developers → your provider → your channel → Messaging API tab.

4. Webhook URL → Edit, paste the Make.com address, click Update. The URL must start with hook. and must not start with dialogflow.cloud.google.com.

5. Verify Webhook and turn Use webhook on.

6. Open LINE Official Account Manager → Settings → Response settings: Webhooks on, Auto-response messages off, Greeting message optional.

Teach Make.com what LINE sends. Right-click module 1 → Run this module only, then send hello to your bot from your phone.

The bubble should show destination and an events list. Module 2 (Iterator) splits events, so every later module reads from module 2: 2.replyToken, 2.source.userId, 2.message.text.

You only need to do this once. If you delete the webhook later, you get a new address and must paste it into LINE Developers again.

The link between module 2 and module 3 carries a filter, Text messages only (2.type = message and 2.message.type = text).

It stops stickers, photos, new followers and LINE’s Verify ping before they reach Dialogflow.

Step 3: Dialogflow connection and Detect an Intent (module 3)

Make.com signs in to your Google Cloud project with an OAuth client that you create. Do Part A once per group, signed in with the Gmail that owns the agent.

Part A: create the OAuth client in Google Cloud

1. Open Google Cloud console and select your agent’s project at the top.

1. APIs & Services → Enabled APIs & services: check that Dialogflow API is listed. If not, open Library, search “Dialogflow API”, click Enable.

1. Google Auth Platform (older consoles: OAuth consent screen) → Get started. App name Make Dialogflow, support email = your Gmail, Audience External, contact email = your Gmail, accept, Create.
2. Audience → add every member’s Gmail under Test users.

3. Clients → Create client: Application type Web application, name Make.com. Under Authorized redirect URIs add exactly:

https://www.integromat.com/oauth/cb/google-custom

4. Click Create and copy the Client ID and Client secret. Wait about 5 minutes before using them.

Part B: connect module 3

1. Open module 3 → Connection → Add → type Make OAuth Google (custom Google connection).
2. Paste Client ID and Client secret → Save → sign in with your Gmail.

3. If Google says the app is not verified, click Advanced → Go to integromat.com (unsafe) → Continue. It is your own app.

4. Pick your project from the list:

Check if your bot in https://dialogflow.cloud.google.com/ already exists and matches with your selected Project ID.

Fill in the fields:

Field	Value	Note
Project ID	pick your project from the list	must match the project shown in Dialogflow settings
Session ID	2.source.userId	already mapped; gives each LINE user their own conversation and contexts
Input Type	Text	already set
Query Input → Text	2.message.text	already mapped
Language Code	en, or th if your agent’s default language is Thai	must match the language your training phrases use

Click Save. Module 3 now returns Query Result › Intent › Display Name, Fulfillment Text and Fulfillment Messages[] for every LINE message.

Step 4: LINE reply modules (7 and 23)

Modules 7 and 23 send replies straight to the LINE Messaging API with an HTTP request. Each one needs your own token.

1. Open module 7 LINE reply: Dialogflow answer → Headers → the Authorization row.
2. Replace YOUR_CHANNEL_ACCESS_TOKEN with your long-lived channel access token in LINE Developers → your provider → your channel → Messaging API tab. Keep the word Bearer and one space in front: Bearer eyJh....

3. Do the same in module 23 LINE reply: rich payload.
4. Leave URL, method, body type and Request content exactly as imported.

What the two bodies do, so you can read them in the run log:

Module	Request content	Sends
7	{"replyToken":"{{2.replyToken}}","messages":[{"type":"text","text":{{6.json}}}]}	the intent’s text answer, made JSON-safe by module 6
23	{"replyToken":"{{2.replyToken}}","messages":[{{21.text}}]}	every LINE custom payload of the intent, joined by module 21

The modules between them are already wired and need no edits:

• 4 Intent and answer stores intent (display name) and reply (Fulfillment Text).
• 19 Each response message walks through Fulfillment Messages[].
• 20 LINE payload to JSON turns each payload.line into JSON; its filter keeps only Platform = LINE.
• 21 Join LINE payloads joins them with commas; its filter drops anything that came out as null or {}.
• 22 Text or rich? sends to module 23 when there is at least one payload. Otherwise its fallback route sends to module 7.

A reply token works once and expires after about one minute, so each message must end in exactly one of modules 7, 23 or 35.

Step 5: Gemini fallback route (modules 34 and 35)

When Dialogflow matches no intent, it returns Default Fallback Intent. The link into module 34 carries the filter Route B: fallback to Gemini (4.intent Equal to Default Fallback Intent), so Gemini runs only then.

If you don’t have an API key, go to https://aistudio.google.com/ and sign in with the account you used to create an agent in Dialogflow as shown in the picture, then click the “Get API key” button.

Then the screen will appear as shown in the picture, press the “Create API key” button.

Then, select the project you use, and then press the Create API key in existing project.

Note: Other Project API keys can be used. It doesn’t have to be the same project as Dialogflow.

Once you get the API key, copy it. Then go back to Make.com, click “Create a connection” button and paste the API key as shown in image. And click “Save” button.

Then, select the Model such as Gemini 3.5 Flash Lite, then select (Model) Input as shown in the image, which will assign the role to this module.

Note: The text can be copied from the content below. You may change this text according to your business information.

“I am an expert in ICDI (International College of Digital Innovation). International College is established in 2011 to incorporate international and bilingual undergraduate programs in Chiang Mai University (CMU). The main academic mission is to facilitate student mobility in ASEAN and Europe. CMU practices ASEAN University Network Quality Assurance (AUN-QA) and ASEAN Credit Transfer System (ACTS) as well as European Credit Transfer System (ECTS). To facilitate the best practices, International College has support roles to other academic faculties in student mobility with credit transfer, Inbound and Outbound visiting professor exchange, Cross Border Co-Research with ASEAN+3 university partners, Common European Framework Reference in English for General Education and Free Elective courses, student leadership program and, other infrastructure and activities for international students in CMU. In CMU, International College plays as the major mechanism for internationalization. Therefore the college assists CMU academic departments to acquire international funding for student and staff exchange in AUN-ACTS, AIMS (ASEAN International Mobility Students), EU-Share and Erasmus Plus as well as Horizon 2020. Other target international scholarships and funds include New Columbo Plan, China Scholarship Council, Reinventing Japan, Korea Foundation, DAAD, Franco-Thai and Newton fund, etc. International College also helps international students to engage the local funds including CMU Presidential Funding, TICA and other Thai government international funding. In 2017, International College has been enhanced her academic roles to entrepreneurship and digital innovation. The enhanced mission directly responds to the Thailand 20 Years Strategic Plan in digital startup especially Digital Economic Cluster for Chiangmai. This aims at development of TransNation Education with leading entrepreneurship and innovation universities in United Kingdom, Australia, China, Korea, etc. The new International College of Digital Innovation (ICDI) offers bachelor, master, doctoral programs in digital innovation and financial technology. Besides the major courses, ICDI also offers General Education and free elective courses in digital entrepreneurship literacy to any CMU students in other faculties. This helps students to learn disruptive digital technologies for engaging new digital economy and society. Today ICDI has 2+2 dual degree programs in entrepreneurship with University of Strathclyde and East China University of Science and Technology. As well as in data analytic, ICDI collaborates with Curtin University and University of Electronic Science and Technology China. To encourage digital startup, ICDI provides academic and training programs for entrepreneurship and innovation. The digital innovation process for SMEs and startups consists of Discovery, Development, Diffuse and Impact by Big Data, mobile application, social network and financial technology respectively. Fundamentally ICDI is doing research and development in cross border e-commerce, logistics and FinTech to bridge Chinese platforms to western platforms for Thai, Chinese and international students. To conduct policy research in entrepreneurship and innovation by using machine learning, research study is based on Supply Side Structural Reform for improving inputs factors for SMEs and startup in new digital economy and society. The intervention of very high-speed broadband Internet such as 4G is the main consideration. The input factors are categorized into land, labor, capital and entrepreneurship and outputs are measured by United Nation Human Development Index which consists of life expectancy, knowledge and net income. By Schumpeterian Cobb-Douglas approach, the creative destruction theory is employed for study of the 10 new S-curve industries of Thailand. The machine learning techniques are used for developing predictive models of decisions, policies, strategies or projects to improve some input factors for the industrial cluster of SMEs and startups. After word, Big Data technology can be utilized for online monitoring the outcome.”

Then, press the Add item button, select the (User) details as shown, then press “Save”.

Module 34 Gemini answer

1. Connection → Add → paste your Gemini API key → Save.
2. AI Model: keep the imported Flash Lite model, or pick another Flash model. Flash models answer within LINE’s one-minute reply window; slower Pro models risk an expired reply token.
3. Messages → item 1 (Role: Model): this is the context Gemini reads before every question. Replace every [BRACKET] with facts about your own platform from Presentation 1. Specific facts (services, hours, prices you are sure of, contact channel) give better answers than general descriptions.
4. Messages → item 2 (Role: User): leave 3.queryResult.queryText. That is the student’s original question.

Module 35 LINE reply: Gemini answer

1. Connection → Add → enter your Channel access token and your User ID (Basic settings tab) → Save.
2. Reply Token: 2.replyToken (already mapped).
3. Messages → Text: 34.result (already mapped).

If you renamed your fallback intent in Dialogflow, type its exact new name in the Route B filter and in the Route A filter (link into module 6), otherwise both routes may run or neither will.

Modules 7, 23 and 35 each have an Ignore error handler attached. A failed reply then stops quietly instead of turning the scenario off, so always check the run log while testing.

Step 6: Build Dialogflow responses for this flow

How you write each intent’s response in Dialogflow decides which module answers in LINE.

Response you build in Dialogflow	Where it appears in module 3	Module that replies	What LINE shows
Text response on the Default tab	Fulfillment Text	7	plain text
Text response on the LINE tab	Fulfillment Text and Fulfillment Messages (Platform LINE, Text)	7	plain text
Custom payload on the LINE tab, {"line": {...}}	Fulfillment Messages (Platform LINE, Payload)	23	any LINE message type
Text and custom payload in the same intent	both	23	payloads only; the text is dropped
Card, Image or Quick Replies built-in types on the LINE tab	Fulfillment Messages, but not as LINE JSON	none	nothing; convert them to custom payloads
No response at all	empty	7	error 400 (text may not be empty)

Rules that follow from the table:

• Give every intent at least one response.
• For rich messages, use LINE tab → Add Responses → Custom payload, one block per message, at most 5 per intent.

• To send text together with a rich message, add the text as one more custom payload: {"line":{"type":"text","text":"Here is our menu"}}.
• Inside "line", use any object from the LINE message types reference. Design Flex messages in the Flex Message Simulator and paste the JSON.

Two custom payload examples:

{"line":{"type":"text","text":"Contact Points","quickReply":{"items":[
 {"type":"action","action":{"type":"message","label":"Phone","text":"phone number"}},
 {"type":"action","action":{"type":"uri","label":"Website","uri":"https://icdi.cmu.ac.th/"}},
 {"type":"action","action":{"type":"location","label":"Share location"}}]}}}

{"line":{"type":"sticker","packageId":"446","stickerId":"1988"}}

Step 7: Test, including the fallback

Test with Run once first, then switch the scenario on. Every message should light up exactly one of modules 7, 23 or 35.

Run these tests in order. For each one, click Run once, send the message from your phone, then check the bubbles and the LINE chat.

Test	Send from LINE	Expected path (bubbles)	Expected in LINE	Pass when
T1	Click Verify in LINE Developers	1 runs, 2 outputs nothing	nothing	Verify shows Success
T2	A phrase of a Default-tab text intent, e.g. hello	1→4, Route A, 19→22, 7	the text answer	7 status 200
T3	A phrase of a LINE-tab text intent	same as T2	the LINE text	7 status 200
T4	A phrase of an intent with one custom payload	Route A, 21 has text, 23	rich message (quick reply, sticker, flex)	23 status 200
T5	A phrase of an intent with 2 or more payloads	same as T4	all payloads in order	23 status 200
T6	Off-topic question in English, e.g. any quiet café near CMU?	Route B, 34, 35	Gemini answer about your platform	35 succeeds, 6 and 7 do not run
T7	Off-topic question in Thai	Route B, 34, 35	answer in Thai	same as T6
T8	Nonsense, e.g. asdfgh qwe	Route B, 34, 35	short polite answer	same as T6
T9	Send a sticker or photo	1, 2, filter “Text messages only” stops	nothing	no error in history
T10	Three different messages within 10 seconds	three separate runs	three answers	each run has one reply

Fallback in depth. For T6 to T8, open the bubbles and check four places:

1. Module 3 output → Query Result → Intent: Display Name Default Fallback Intent, Is Fallback true. Note Intent Detection Confidence.
2. Filter labels on the router: Route A shows ⊘0, Route B shows a tick and 1.
3. Module 34 output → Result: the text Gemini wrote. Check it uses facts from your prompt, not invented ones.
4. Module 35: no error. If it fails with an invalid reply token, the run took too long; check the run duration in History and choose a faster model.

Then tune the boundary between your intents and Gemini:

You see	Meaning	Change
A real question about your platform goes to Gemini	the intent lacks matching training phrases	add 5 to 10 more phrases to that intent, including Thai if users write Thai
An off-topic question triggers one of your intents	the match threshold is too low	Dialogflow → gear → ML Settings → Classification threshold: raise it (for example from 0.3 to 0.5), save, retest
Gemini invents facts	the prompt lacks those facts	add them to module 34’s Model message, or tell Gemini to say it is not sure
Gemini answers in English to Thai questions	the prompt rule is missing	keep “Reply in the same language the user writes in” in the prompt

Troubleshooting

What you see	Where	Cause	Fix
Bot never answers, no new run in history	Make.com	Scenario off, or LINE webhook still points to Dialogflow	Turn scheduling on; set the Make.com URL in LINE Developers
Module 1 output shows queryResult, not events	1	Make.com URL pasted into Dialogflow Fulfillment instead of LINE	Create a new webhook, put it in LINE Developers only
Iterator (2) outputs nothing	2	Data structure not learned	Right-click 1 → Redetermine data structure, send a message
redirect_uri_mismatch	Google sign-in	Redirect URI differs	Add https://www.integromat.com/oauth/cb/google-custom exactly
“has not completed the Google verification process”	Google sign-in	Gmail not a test user	Publish the app, or add the Gmail as a test user
Status 400 when saving the connection	3	Wrong secret or client type	Copy the secret again; type must be Web application
Detect an Intent: 403 or invalid_grant	3	Sign-in expired, wrong project	Reauthorize the connection; check Project ID
Every message goes to Gemini	3, 5	Language code differs from agent, or weak training phrases	Match en/th; add training phrases
LINE 400 May not be empty, messages	23	No real payload reached module 21	Check the intent has a LINE custom payload in {"line":{...}} form
LINE 400 text may not be empty	7	Intent has no text response	Add a response to the intent
LINE 400 Invalid reply token	7, 23, 35	Two routes replied, or reply came after about 1 minute	Check filters; use a Flash model
LINE 401	7, 23	Bearer missing or token wrong	Header = Bearer + space + token
Gemini answers for intents that should match	5	Route B filter missing	Filter on link into 34: 4.intent Equal to Default Fallback Intent

Modules with an Ignore handler fail quietly. Open History, click the run, then click the module bubble and read Status code and Data.