Join the Community

Workflow Issues: HTTP API Status Code Logic & AI Field Mapping Troubles | Clay

Workflow Issues: HTTP API Status Code Logic & AI Field Mapping Troubles

·

·

Hi everyone, I'm currently experiencing two issues in my workflow that I can't resolve on my own: 1. HTTP API status code logic isn't working as expected I have working HTTP API steps for four different URLs (Impressum, Kontakt, Über uns, About). Each step returns a statusCode, and I’m trying to build a conditional formula like: js Kopieren Bearbeiten {{Step_1_HTTP_API_Impressum_URL.statusCode}} == 200 ? {{Impressum_URL}} : {{Step_2_HTTP_API_Kontakt_URL.statusCode}} == 200 ? {{Kontakt_URL}} : ... But Clay doesn’t seem to evaluate these conditions correctly. I’ve tried various formats, but the logic either fails or doesn’t trigger correctly. Here is the Loom video for the HTTP API problem: https://www.loom.com/share/07c3bc43cbeb490c81dbb4e5eb9e2464?sid=72656552-7d86-4f42-8b1c-d85c322f397b 2. AI output fields are not mapping to columns My AI Web Researcher returns JSON like this: json Kopieren Bearbeiten { "Ansprechpartner_I_Vorname": "Sercan", "Ansprechpartner_I_Nachname": "Toku" } These field names exactly match existing columns in my table. I’ve configured the output format as "Fields", and I’ve added clear “Help AI” descriptions to both target columns. Despite that, the extracted values are not populating the columns. The preview shows the correct values in the reasoning, but the actual columns remain empty. Here is the Loom video for the mapping: https://www.loom.com/share/6b03b48417e044c4a67eb6036b92db72?sid=41dcc933-1e28-4c9e-8de6-c389c90c4ac9 Any help would be appreciated!

12 comments

Clay
APP
·
·
For your HTTP API conditional logic issue, the problem likely stems from JSON formatting or syntax errors in your formula structure. When using dynamic variables in conditional formulas, ensure proper quotation marks around string values and correct comma placement between conditions. For your AI output mapping issue, this often occurs due to JSON formatting problems even when field names match exactly. Common causes include missing quotation marks around keys, trailing commas, or incorrect data types in the JSON structure. Could you copy and paste both your conditional formula code and the exact JSON output from your AI Web Researcher so I can identify the specific formatting issues?
Sources: HTTP API integration overview | Documentation | Clay University HTTP API integration overview | Documentation | Clay University FAQ: Why am I seeing the "failed to parse body" error when using the HTTP API enrichment?
Philip H.
·
·
Sure. Here is the table: https://app.clay.com/workspaces/670253/workbooks/wb_0syvllqKCpAiGTkYu9U/tables/t_0syvllqF7n7jrGyYrEx/views/gv_0syvllqk4oC3YUccMmC
Clay
APP
·
·
Hello! Thanks for reaching out. Curious to know a few things while I troubleshoot this: - Are you trying to create a waterfall? If status code is 200 for Step 1 do not run the next steps? If yes, what we can do is pull out the status column for each API column and if for one row it is successful in Step 1, it breaks the flow and we can merge the results in a last merged column. This way irrespective of which step it is successful in, the result will be displayed in the last column. You mentioned in the query that if Step 1 HTTP API is successful then use that URL so on and so forth but in the run condition setup for the steps, I can see the following conditions:
(191 kB)
Can you tell me the difference between the prompt you put in the chat vs these run conditions? Trying to understand the entire objective before suggesting a solution!
Clay
APP
·
·
For your second query, taking a look now!
Clay
APP
·
·
When it comes to mapping out columns for first name and last name, the JSON output will always create new fields. It will not map columns simply because you gave it more context in the setup of the AI column. To map it you can merge the results of the two columns into a third column. I setup a column for you to checkout.
(107 kB)
Let me know!
Philip H.
·
·
Hi, thanks for your response! I'm trying to build a workflow in Clay that checks whether contact data is missing, and if so, tries to scrape the correct webpage from a given company website. Here's what I want to achieve step-by-step:
1.
Run the automation only if the contact has:
No first name
No last name
But a website listed in the column website_claim
2.
If these conditions are met, I want to:
Try to check if website_claim/impressum exists
If that doesn’t exist (status code ≠ 200), then try /kontakt
If that also doesn’t work, then try /about
As soon as I find one that exists (returns status 200), I want to scrape that page
I've tried using the AI Prompt Builder for this, but I couldn’t get conditional logic working with status codes (like if status == 200). I understand that Clay doesn’t support ternary expressions directly, but maybe there’s a workaround using helper columns? I tried to use this {{ Step_Impressum.statusCode == 200 ? Step_Impressum.url : null }}, but it did not work. Can you suggest the best way to model this logic in Clay? As for the new name columns: 1. Column Mapping Doesn’t Apply to All Rows I tried updating the result columns to Resultat_Vorname and Resultat_Nachname. While this mapping worked for the first two rows, it didn’t apply to later rows when I ran the process again. Can you help me understand why the column mapping doesn’t consistently take effect? I have made a video to illustrate this: https://www.loom.com/share/15e44d472c11492f8b039b0b85485354 2. Name Extraction Fails on Some Rows In a later row (which I show in the video), the system failed to extract a name — even though:
I highlighted the name clearly in the page content, and
showed the regex I used.
Why does this happen, and how can I fix it? 3. Prompt Language: English or German? Some prompts fail when written in German, and others work better in English — even though the content is in German. What’s best practice: Should I write prompts in German (to match the input language), or in English (to match the system’s base)? 4. coalesce Formula Doesn’t Prioritize Correctly Even though I used the coalesce function as advised, the result doesn't seem to:
pick the first working URL (200 status) among impressum, kontakt, or about
or apply conditional logic correctly based on status code.
I’m trying to scrape the first valid page, based on availability. What’s the best way to implement this logic using HTTP API checks? 5. Column Workaround Acceptable (if needed) If the result can’t be reliably written to Resultat_Vorname and Resultat_Nachname directly, I’m okay with a workaround like:
writing to temp columns first
then copying into the target columns as a second step — as long as this works reliably
Thanks in advance!
Bo (.
·
·
I recorded this walkthrough video showing exactly how to set this up: https://cln.sh/xHzW9RTJ Let me know if you have any follow-up questions!
Philip H.
·
·
Hi Bo! Thanks for your help! Your tip (To use the "Create or modify content" use case worked well! I do have some questions about the use of FindYMail and other such email finders, as outlined here: https://www.loom.com/share/ba3038a265224061b4d043b74092e25c?sid=c2f7fded-1bab-455f-8a98-67ae8c7dc387 Here is the table, should you need it: https://app.clay.com/workspaces/670253/workbooks/wb_0syvllqKCpAiGTkYu9U/tables/t_0syvllqF7n7jrGyYrEx/views/gv_0syvllqk4oC3YUccMmC Thanks for your help!
Bo (.
·
·
I recorded this walkthrough video showing exactly how to set this up: https://cln.sh/SJLBG6cM Let me know if you have any follow-up questions!
Philip H.
·
·
Hi Bo, thank you for the video – super helpful! Do you happen to know if there’s a reliable way to extract direct phone numbers (ideally personal extensions) of decision-makers, similar to how Findymail helps with personalized emails? We're trying to avoid collecting just central office numbers or generic contact details like info@ and we are in Germany, so I would also like to understand if tools like Apollo.io can also find German numbers. We're specifically looking for the direct phone extension of the decision-maker.. Thanks in advance for any tips you might have!
Daniel K.
·
·
Hey there, have you tried using our mobile phone waterfall? It’s designed to surface direct dials and mobile numbers specifically for decision-makers.
(244 kB)
If there's anything else I can assist you with, please let me know! 😊 Have a great Thursday, Daniel
Philip H.
·
·
Yes, this issue is resolved. I asked how the cohorts program works and how we get the assignments in the cohorts channel.