I have a list of restaurants in a PDF file I've uploaded and associated with my agent. However, whenever I ask the agent to give me nearby restaurants, it gives me the descriptions of options but skips saying the actual names of the places. Org ID: 87b7f212-928b-4f5f-acf9-d2ab17ed8f59 Agent ID: e6dd0cbb-672b-4240-8d5a-d34555163885 Call ID: 2fb681f5-c5e0-45e5-8663-ebfb5f17bec2 Actual Output: "Oh, absolutely. We have a few great pizza places near the resort. Let me see. Okay. I'm looking at my list here. There's which is just half a mile away. It's a lively bar with live music, craft beers, and, of course, pizza. It's a popular spot for locals, and has a great nightlife scene. There's also, which is about a quarter mile away, it's a laid back beach bar and grill with burgers, fish tacos, and frozen drinks. They also have a great pizza menu. Would either of those work for you or are you looking for something a little more specific?" (notice that it just skips saying the name of either of the recommendations it talks about) Expected Output: "Oh, absolutely. We have a few great pizza places near the resort. Let me see. Okay. I'm looking at my list here. There's Chicho's Pizza Backstage, which is just half a mile away. It's a lively bar with live music, craft beers, and, of course, pizza. It's a popular spot for locals, and has a great nightlife scene. There's also The Shack, which is about a quarter mile away, it's a laid back beach bar and grill with burgers, fish tacos, and frozen drinks. They also have a great pizza menu. Would either of those work for you or are you looking for something a little more specific?" (bolded the names it should have said) I've also attached a screenshot of the PDF file that I have associated with the agent, in case it's helpful. Any ideas why the restaurant/bar names don't show up in the transcript output? https://cdn.discordapp.com/attachments/1357014213800759487/1357014214509592596/Screenshot_2025-04-02_at_11.28.43_AM.png?ex=67eea978&is=67ed57f8&hm=2caf3fd23df48417e95515653ad8b83996073d62bdce2a1b96678bbed241d242&

Hi there! 👋 I see you're having an issue with your voice AI agent skipping restaurant names when reading from your PDF file. I'm unable to diagnose exactly what's happening with your specific call, but I can definitely help with some troubleshooting steps based on similar issues others have experienced. Here are several potential solutions to fix the missing restaurant names: 1. **Check Transcriber Settings**: The issue might be with the transcriber. According to [troubleshooting documentation](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1331353180570914878), names are often challenging for transcribers like Deepgram's Nova 2 model to capture correctly. 2. **Enable Word Boosting**: Make sure word boosting is enabled and consistently applied across all transcription processes. This can significantly improve name recognition according to [support information](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1336366437773738150). 3. **Adjust Endpointing Settings**: The default Deepgram endpointing value of 10ms might be too aggressive. Try increasing it to 300ms as [suggested in troubleshooting guides](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1343636482874933340). 4. **Modify startSpeakingPlan.waitSeconds**: Set this parameter to 0.8 seconds to give the model more time to recognize and capture names correctly, as [recommended in support documentation](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1333510134617604147). 5. **Consider Document Formatting**: Ensure your PDF is in a format that's easily parsed. Sometimes complex formatting can cause issues with name recognition. Breaking the document into smaller sections might help, according to [troubleshooting information](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1329700156827893821).

I just tried to troubleshoot this by creating another agent that has my list of restaurants/bars/events directly placed in the system prompt (attached screenshot). Even then, it still just ignores saying the actual names of the restaurants? Agent ID: e07e906d-aa6d-443c-95e8-8c47f9299189 Call ID: edbcbc0a-4898-40b7-9076-3406347e7fa1 Actual Output: "Hi there. Yes, we have 2 fantastic on-site restaurants for you to enjoy during your stay. First, we have, which offers fresh seafood, handcrafted cocktails, and stunning ocean views. It's a great spot for breakfast, lunch, or dinner. Second, there's on specializing in modern Mexican dishes like tacos and enchiladas. Served in a casual and relaxed atmosphere. Both restaurants are located right here at the resort, so there's super convenient for you. Let me know if you'd like any additional details or help with reservations." Expected Output: "Hi there. Yes, we have 2 fantastic on-site restaurants for you to enjoy during your stay. First, we have Oceans 14, which offers fresh seafood, handcrafted cocktails, and stunning ocean views. It's a great spot for breakfast, lunch, or dinner. Second, there's Oceans Ole specializing in modern Mexican dishes like tacos and enchiladas. Served in a casual and relaxed atmosphere. Both restaurants are located right here at the resort, so there's super convenient for you. Let me know if you'd like any additional details or help with reservations." (bolded the names that should have been spoken by the agent, but were not) Thanks in advance for your help.

checking if this is resolved/solved for you?

@Shubham Bajaj nope still a problem, for some reason the voice agent won’t say the names of the restaurants

looking into it.

@Shubham Bajaj Looking at the documentation for chunk plans (https://docs.vapi.ai/assistants/voice-formatting-plan), it doesn't look like any step removed names. Would you mind letting me know which step is responsible for removing the names? And is it possible to have the LLM output the names to speak, while keeping rest of the chunk plan steps?

Hey CrispyChips, There are a few utility functions (such as Markdown and Asterisks) which are responsible for trimming the Markdown-related content. Add the following formatter enabled to your account, and this issue won't appear again for you.

"chunkPlan": {
           "enabled": true,
           "formatPlan": {
             "enabled": true,
             "formattersEnabled": [
               "acronym",
               "dollarAmount",
               "email",
               "date",
               "time",
               "distance",
               "unit",
               "percentage",
               "phoneNumber",
               "number"
             ]
           }
         }
       }

Thanks, I'll give this a try and let you know if I have any issues.

Marking this ticket as Solved ✅