Skip to content

Commit 5bd6c8b

Browse files
geemi725geemi725
committed
feedback jwu
1 parent d9de641 commit 5bd6c8b

3 files changed

Lines changed: 3 additions & 48 deletions

File tree

‎chapter_3/agent_implementation.ipynb‎

Lines changed: 1 addition & 46 deletions
Original file line numberDiff line numberDiff line change
@@ -354,7 +354,7 @@
354354
"\n",
355355
"We'll provide a short fragment of its sequence so the notebook runs quickly. In a real workflow, you'd paste the full sequence from UniProt.\n",
356356
"\n",
357-
"Let's use a truncated fragment of human EGFR (from UniProt P00533). In practice, use the full sequence from https://www.uniprot.org/uniprot/P00533\n"
357+
"In this example let's use a truncated fragment of human EGFR (from UniProt P00533). You can find the full sequence from https://www.uniprot.org/uniprot/P00533\n"
358358
]
359359
},
360360
{
@@ -401,51 +401,6 @@
401401
"\n",
402402
"Note that you can use the RAG pipeline we learned in section 2.2 as a tool here to extract literature data."
403403
]
404-
},
405-
{
406-
"cell_type": "markdown",
407-
"id": "41499148",
408-
"metadata": {},
409-
"source": [
410-
"## 3.2.7 Standard approaches to deploying agents \n",
411-
"\n",
412-
"Anthtopic has released a open-source protocol known as Model Context Protocol (MCP) to standardize the these practices. Read more [here](https://www.anthropic.com/news/model-context-protocol).\n",
413-
"\n",
414-
"Imagine you want your AI agent to use 20 different tools such as search PubMed, query UniProt, read files, run BLAST, write to a database. If you used the notebook approach, you'd have to manually write and wire up all 20 functions yourself, in every project, every time.\n",
415-
"Instead MCP introduces a plug-and-play architecture with two distinct roles:\n",
416-
"\n",
417-
"**MCP Server**\n",
418-
"\n",
419-
"- A standalone program that owns and exposes tools. For example: a \"UniProt MCP Server\" that knows how to fetch protein data from UniProt's API\n",
420-
"- It advertises what tools it has, what inputs they take, and runs them when asked\n",
421-
"- Anyone can build one, and they can be shared and reused across projects\n",
422-
"\n",
423-
"**MCP Client**\n",
424-
"\n",
425-
"- The AI agent (or the app hosting it) that connects to one or more servers\n",
426-
"- It discovers what tools are available, and calls them when the model decides to\n",
427-
"- Claude.ai, for example, acts as an MCP client when you connect it to Google Drive or Slack\n",
428-
"\n",
429-
"\n",
430-
"\n",
431-
"The relationship looks like this:\n",
432-
"\n",
433-
"```\n",
434-
"[ Your AI Agent / Claude ] ← MCP Client\n",
435-
" │\n",
436-
" │ \"What tools do you have?\"\n",
437-
" ▼\n",
438-
"[ UniProt MCP Server ] ──→ tool: fetch_protein()\n",
439-
"[ PubMed MCP Server ] ──→ tool: search_papers()\n",
440-
"[ BLAST MCP Server ] ──→ tool: run_alignment()\n",
441-
"```\n",
442-
"\n",
443-
"You can see a more detailed explanation and an implementation in chapter 3.4. In the meantime you can read more about MCP here:\n",
444-
"\n",
445-
"- [what is an MCP?](https://www.youtube.com/watch?v=eur8dUO9mvE)\n",
446-
"- [MCP vs API](https://www.youtube.com/watch?v=7j1t3UZA1TY)\n",
447-
"- [Concepts of MCP](https://modelcontextprotocol.io/docs/learn/architecture)"
448-
]
449404
}
450405
],
451406
"metadata": {

‎chapter_3/agent_mcp_example.ipynb‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -6,7 +6,7 @@
66
"source": [
77
"# 3.4 Implement an Agent Using Claude Agent SDK\n",
88
"\n",
9-
"In chapters 3.2 and 3.3 we saw how to write boilerplate code and implement agents. But the problem is, if you are not a python programmer you mind find this to be tedious. As a scientist this may not be the problem you want to solve. \n",
9+
"In chapters 3.2 and 3.3 we learned how to write boilerplate code and implement agents. But the problem is, if you are not a python programmer you might find this to be tedious. This may not be the main problem you want to solve and might be interested in a simpler solution.\n",
1010
"\n",
1111
"## 3.4.1 Claude Agent SDK\n",
1212
"\n",

‎chapter_3/agents.md‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -81,7 +81,7 @@ Key Concepts to Remember
8181

8282
At FutureHouse, we're seeking to accelerate scientific research with an AI agents designed to automate many of the critical steps on the path toward scientific progress. We have built a series of AI agents specialized for tasks including information retrieval, information synthesis, chemical synthesis design, and data analysis.
8383

84-
As introduced in chapter 2, [PaperQA2](https://github.com/Future-House/paper-qa) is our scientific information retrieval system. Our data analysis agent is named [Finch](https://github.com/Future-House/finch). Finch is an AI agent framework designed to perform complex scientific data analysis tasks by iteratively working through Jupyter notebooks. This agent takes in datasets and prompts, then systematically explores, analyzes, and interprets the data to provide comprehensive answers and insights. Then we have [Robin](https://www.futurehouse.org/research-announcements/demonstrating-end-to-end-scientific-discovery-with-robin-a-multi-agent-system), our first multi-agent system for scientific discovery. We applied Robin to identify ripasudil, a Rho-kinase (ROCK) inhibitor clinically used to treat glaucoma, as a novel therapeutic candidate for dry age-related macular degeneration (dAMD), a leading cause of irreversible blindness worldwide. The second iteration of Robin, named [Kosmos](https://edisonscientific.com/articles/announcing-kosmos?gad_source=1&gad_campaignid=23563065812&gbraid=0AAAABB7BYdBNoPj2BU82YRgDuN7FLSbFp&gclid=CjwKCAjwwJzPBhBREiwAJfHRnZ_daxcLWR4IyY7swufHvUA5GBDm-dNoDpG-gSJng_9pe96pNd0ciBoCT7wQAvD_BwE). The core innovation in Kosmos is the use of structured world models, which allow efficient incorporation of information extracted over hundreds of agent trajectories. This also allows to maintain coherence towards a specific research objective over tens of millions of tokens.
84+
As introduced in chapter 2, [PaperQA2](https://github.com/Future-House/paper-qa) is our scientific information retrieval system. Our data analysis agent is named [Finch](https://github.com/Future-House/finch). Finch is an AI agent framework designed to perform complex scientific data analysis tasks by iteratively working through Jupyter notebooks. This agent takes in datasets and prompts, then systematically explores, analyzes, and interprets the data to provide comprehensive answers and insights. Then we have [Robin](https://www.futurehouse.org/research-announcements/demonstrating-end-to-end-scientific-discovery-with-robin-a-multi-agent-system), our first multi-agent system for scientific discovery. We applied Robin to identify ripasudil, a Rho-kinase (ROCK) inhibitor clinically used to treat glaucoma, as a novel therapeutic candidate for dry age-related macular degeneration (dAMD), a leading cause of irreversible blindness worldwide. The second iteration of Robin, named [Kosmos](https://edisonscientific.com/articles/announcing-kosmos?gad_source=1&gad_campaignid=23563065812&gbraid=0AAAABB7BYdBNoPj2BU82YRgDuN7FLSbFp&gclid=CjwKCAjwwJzPBhBREiwAJfHRnZ_daxcLWR4IyY7swufHvUA5GBDm-dNoDpG-gSJng_9pe96pNd0ciBoCT7wQAvD_BwE), is our first AI scientist. The core innovation in Kosmos is the use of structured world models, which allow efficient incorporation of information extracted over hundreds of agent trajectories. This also allows the agent to maintain coherence towards a specific research objective over tens of millions of tokens.
8585

8686

8787
## 3.1.4 Additional Reading

0 commit comments

Comments
 (0)