Module 3 — Dashboard
Lesson Title
The LangSmith Deployment Dashboard
Goal
Orient the student to the LangSmith dashboard so they can find their deployment, inspect runs and traces, and monitor production behavior. This module is delivered as a video walkthrough; the agent's role is to answer follow-up questions about what the student saw.
How to run this lesson
- This module is a video walkthrough — there is no companion script to run.
- Ask what the student wants to find in the dashboard (a specific trace, a deployment URL, runs by a user, cost metrics, etc.) and describe where to look.
- If the student asks "where do I see X?", walk them to the section of the dashboard rather than explaining concepts in the abstract.
- Reinforce the connection to other modules: a trace shows the same
thread_id/assistant_id they used in module 2; an env-var screen is where they confirm the .env from module 1 actually reached the deployment.
Key concepts to cover
- The dashboard URL —
https://smith.langchain.com/; the deployments list is the entry point
- A deployment page — graph list, deployment URL, env vars, revisions, traffic
- Traces — one row per run; expand to see the full agent trajectory, tool calls, and inputs/outputs
- Filtering traces by
thread_id, assistant_id, or metadata.owner — the same identifiers the SDK uses
- Monitoring — request rate, error rate, latency, token usage, cost
- Revisions — each deploy is a new image; you can roll back if a deploy breaks
- Env vars — what's actually packaged into the deployment (vs what's only in your local shell)
- The relationship between dashboard objects and SDK calls —
client.threads.search() and the threads tab show the same data
Tone guidance
Concise and practical. The student is not learning new APIs in this module — they're learning where things live. When they ask about a feature, answer with a navigation path ("click your deployment → Traces → filter by …") rather than a conceptual explanation.
Reference material
Full reference material is in information.md in this directory. Read it before answering factual questions.
1---2name: module-33description: Teaching instructions for Module 3 (Dashboard) — use when module_id is module-34---56# Module 3 — Dashboard78## Lesson Title9The LangSmith Deployment Dashboard1011## Goal12Orient the student to the LangSmith dashboard so they can find their deployment, inspect runs and traces, and monitor production behavior. This module is delivered as a video walkthrough; the agent's role is to answer follow-up questions about what the student saw.1314## How to run this lesson15- This module is a video walkthrough — there is no companion script to run.16- Ask what the student wants to find in the dashboard (a specific trace, a deployment URL, runs by a user, cost metrics, etc.) and describe where to look.17- If the student asks "where do I see X?", walk them to the section of the dashboard rather than explaining concepts in the abstract.18- Reinforce the connection to other modules: a trace shows the same `thread_id`/`assistant_id` they used in module 2; an env-var screen is where they confirm the `.env` from module 1 actually reached the deployment.1920## Key concepts to cover211. The dashboard URL — `https://smith.langchain.com/`; the deployments list is the entry point222. A deployment page — graph list, deployment URL, env vars, revisions, traffic233. Traces — one row per run; expand to see the full agent trajectory, tool calls, and inputs/outputs244. Filtering traces by `thread_id`, `assistant_id`, or `metadata.owner` — the same identifiers the SDK uses255. Monitoring — request rate, error rate, latency, token usage, cost266. Revisions — each deploy is a new image; you can roll back if a deploy breaks277. Env vars — what's actually packaged into the deployment (vs what's only in your local shell)288. The relationship between dashboard objects and SDK calls — `client.threads.search()` and the threads tab show the same data2930## Tone guidance31Concise and practical. The student is not learning new APIs in this module — they're learning where things live. When they ask about a feature, answer with a navigation path ("click your deployment → Traces → filter by …") rather than a conceptual explanation.3233## Reference material34Full reference material is in `information.md` in this directory. Read it before answering factual questions.
langchain-ai