Supply Drop - Medium

This is the AnnaSeva site. This link opens it on your session's level, already chosen. Sign in if you are asked to, then check the dashboard here after your run.

https://pathfinder.automationanywhere.com/challenges/training/Build_An_Agent_Resources/annaseva/auth/picker.html?tier=medium

This is the link to the control room you will be performing all hands on activities within.

https://aa-community.cloud.automationanywhere.digital/

Your session's build is in Bots/BAA/AnnaSeva/Medium_Chain Reaction.

The page below has all the API documentation for AnnaSeva.

View API Documentation

Steps:

1. Navigate to Medium_Chain Reaction/AG_Resolution/AT_SupplierHistory and click the name to open it. Its input iStrApiKey and its output oStrSupplierHistoryJson are already there. Notice it has no actions yet!
2. In Actions, find REST Web Services and drag its Get method onto the canvas.
3. Set URI to:

   https://vvnvfidppaulgbusrqta.supabase.co/functions/v1/super-endpoint/supplier-history

4. Leave Authentication mode on No authentication.
5. Under Custom headers, press Add header. Name it X-API-Key, choose Insecure string for the value and paste:

   $iStrApiKey$

   • Press Add.
6. It warns the value is not encrypted: that is expected, it only passes on the key the process was given.
7. Press Add header again and name it apikey. Keep Credential, click Choose, and pick the locker BuildAnAgent, the credential AnnaSeva and the attribute AnnaSevaAPI. Press Confirm, then Add.
8. Tick Capture failure response. Under Assign the output to a variable, choose Dictionary and pick dictSup.
9. Drag String: Assign below it. Set the source string to:

   $dictSup{Body}$

   • and the destination to oStrSupplierHistoryJson.
10. Press Save and close the task.

Steps:

1. Navigate to Medium_Chain Reaction/AG_Resolution/AT_CascadeImpact and open it. It already reads the policy. Take note: the AI Skill that works out what each option leads to is not called yet.
2. In Actions, find AI Skill and drag Execute below the last action.
3. Click Choose, then the Browse button at the top right, and get Medium_Chain Reaction/AG_Resolution/AS_CascadeImpact from there, in private.
4. Its inputs appear in this order. Set policy to:

   $dictPolicy{Body}$

5. Set quote to:

   $iStrQuoteJson$

6. Set situation to:

   $iStrSituation$

7. Assign the response to oStrImpactJson and the error to oStrSkillError.
8. Press Save and close the task.

Steps:

1. Navigate to Medium_Chain Reaction/AG_Resolution/AG_Resolution and open it. Its role and seven tools are already there. Go to Tools.
2. Press Add tool, select API Task and click Choose, then the Browse button at the top right (not Search), and get Medium_Chain Reaction/AG_Resolution/AT_SupplierHistory from there, in private. In its Tool description for AI, clear any text and paste:

   Supplier: every past delivery with what was promised, what arrived, how late, and its cycle (normal, festival or surge). Use it when a line depends on a supplier's load, to judge whether that load can be counted on today. Judge the supplier on the deliveries that look like today, not on the average.

3. Press Add tool, select API Task and click Choose, then the Browse button at the top right (not Search), and get Medium_Chain Reaction/AG_Resolution/AT_CascadeImpact from there, in private. In its Tool description for AI, clear any text and paste:

   Cascade impact: works out what each way of settling a decision sets off - firm and conditional meals, cost and premium, what rides on the supplier and who must approve - from the situation you describe in plain sentences and the emergency quote exactly as it was returned. It works from the facts and the quote, and the decision you write rests on its answer.

4. Take note: no tool says when to call it in a fixed order. Each says what it is for, and the agent picks what to ask next from what the last one told it.

Steps:

1. Go to the Goal and paste:

   Leave the shift in a state the Operations Head can decide from in two minutes. Every shortfall you are allowed to close is closed. Every open supplier ticket is dealt with. Anything only the Operations Head can decide, such as how much of a relief order to promise firm when the shelf alone cannot cover it, is framed as one decision with two costed options, the evidence behind each and the district figures for each, so the approval form can show it and the process can carry out whichever option they pick. You are done when that decision is ready and every output is filled as its description asks.

2. Go to the Action plan and paste:

   Work like a careful planner, not a checklist. Call one tool at a time and let what each one tells you decide what you ask next.
   
   What counts as stock: only usable stock and loads already confirmed at the gate (policy S6). Every line on the gap list is short even after every confirmed load has been counted: close each one you have the authority to close. The gap list is not the whole job, though: it leaves out every line that a load still to arrive is covering, and on a relief day those are the risk. The district's relief orders are the plan's P4 lines. The inventory tool's load check sets each load still to arrive against the line it feeds and its usable stock alone. Where the shelf cannot cover the need, the line rides on that load: it is not short on paper, but it is a risk to understand, not a shortfall to fix. When a relief order rides on a load, it is the Operations Head's decision however reliable the supplier looks: judge the supplier for the evidence, price the alternative and frame the decision.
   
   When something depends on a supplier, judge that supplier on deliveries like today's. A relief surge is not a normal day, so a good average says little. What they did on their surge orders, and what they have written to us, says a lot.
   
   Never judge a kitchen by its totals alone: split its production by programme. The youngest children's meals (P2) are never shorted (S1.1), and a kitchen that has been shorting them is not the place to send more work.
   
   Know whose call it is (S8). Moving or substituting stock inside policy is yours: take the cheapest legal fix and stop. Inside policy means a move only on a lane S5 approves, from a kitchen whose spare in the inventory tool's staple check covers it (S4 keeps its day plus one production day's buffer). How much of a relief order to promise firm, like any emergency purchase, belongs to the Operations Head. A supplier ticket is routine work and never an escalation: deal with each one the way the triage says.
   
   A decision you frame carries the priced alternative and what each option sets off. Every number you write comes from a tool, or from arithmetic on a tool's numbers. Never invent a figure, a price or a reference.
   
   Nothing irreversible is yours. You never notify the district, buy, approve or file the escalation: the process does each of those after the Operations Head decides.
   
   Every tool answers even when it refuses. Read each reply for ok or error before you rely on it, and when something is refused, read why and find another legal way.
   
   Pass the API key to every tool exactly as you were given it, and never write it anywhere else.
   
   Your outputs feed the escalation, the Operations Head's form and the steps that carry out their choice. Fill each one exactly as its description asks.

3. Take note: there are no numbered steps and no answers in it. It says what counts as evidence, whose call each fix is and what is never the agent's to do.
4. Press Save and close the agent.

Steps:

1. Navigate to Medium_Chain Reaction/PR_AnnaSeva_ChainReaction and open it.
2. Between CheckMail and Escalate, remove the connecting line.
3. Drag an AI Agent from the Elements list onto the canvas, between them. Set the Element ID to Resolution.
4. Click Choose, then the Browse button at the top right, and get Medium_Chain Reaction/AG_Resolution/AG_Resolution from there, in private.
5. Under Task assignment, choose Role and click Choose. Type AAE_Robotic Interface User in the search box and press Enter, then select it and click Choose.
6. Connect CheckMail to Resolution, and Resolution to Escalate: drag from the dot at the bottom of one step to the dot at the top of the next.
7. Its input values appear in this order. Set iStrApiKey to:

   $iStrApiKey$

8. Set iStrGapsJson to:

   $&FindGaps{output}{oStrGapsJson}$

9. Tick iStrMailJson (it starts unticked) and set it to:

   $&CheckMail{output}{oStrMailJson}$

10. Set iStrPlanJson to:

    $&CaptureWork{output}{oStrPlanJson}$

11. Press Save.

Steps:

1. Navigate to Medium_Chain Reaction/FM_OpsHeadApproval and open it. Press Preview: a plain grey form.
2. Go to the Styles tab, select everything in the editor, and paste:

   The Ops Approval Queue form code - click to copy

3. Press Preview again. It is now the Ops Approval Queue. Preview shows the layout only, so the options do not respond here: in Action Seven you pick one, and the form warns you what it will do before you record it.
4. Press Save and close the form.
   • The form is already in the process: its step, OpsHeadApproval, sits between Recommend and RecordDecision and is filled from the run. Nothing to change there.

Steps:

1. Open the AnnaSeva link at the top of this page. On the dashboard, press Reset at the top right and press OK, so the day starts fresh.
2. Navigate to Medium_Chain Reaction/PR_AnnaSeva_ChainReaction, open it and press Run.
3. Get your key: on the AnnaSeva dashboard, press Settings at the top right, then Generate / regenerate API key, and copy the key it shows. It starts aa_live_ and is shown only once. Paste it into iStrApiKey and click Confirm.
4. Watch it in Co-Pilot. It takes about 7 to 8 minutes before the form arrives, and about 5 more after you record your decision. The agent picks each next step from what the last tool returned.
5. When the form arrives, read the rule card, then the brief and the recommendation. The brief's THE LOAD line names the delivery the conditional meals depend on, and its status as Inbound Deliveries shows it. AnnaSeva's policy S6 (Policy Viewer on the dashboard) says today: day plans may count only deliveries with status CONFIRMED, and in-transit quantities are a risk flag, not stock. So CONFIRMED means Option 1; anything else means Option 2.
   • The recommendation applies the same rule to that line. If it ever says otherwise, choose by the rule: the rule decides, not the agent.
6. The reason box arrives filled with the recommendation's reason. Keep it, or write your own of at least 20 characters naming the load, its status and the policy. Type your name and press Record decision.
7. Open the AnnaSeva dashboard: click Back to systems on the Settings page (or All systems on any other AnnaSeva page), then AnnaSeva Ops. The shift status reads complete, 4 of 4.